How bad is MP3 really? [Archive] - The Art of Sound Forum

dave2010

22-04-2010, 07:48

Actually I really want to extend this not only to MP3, but other forms of lossy compression, such as AAC, aac+ (AAC-HE) etc.

Some round here have written that MP3 is the work of the devil, but I have to say that for recordings done at above 220 kbps I often find it hard to tell the difference between MP3 and CD. I find that with AAC a lower bit rate is acceptable. I am aware that aac+ can give OK-ish results with very low bit rates, but it's not high quality.

Also, don't some of us listen to DAB radio? That's Musicam/MP2, and the best we can hope for is 192 kbps from Radio 3. It's sometimes bearable - though not always.

My point is that can we really assert that we should never use lossy compression, or never listen to files which are compressed? Sometimes I've thought there were problems with MP3, but then the same problems seem to be on the original CDs. If the microphones were stuck in the wrong place, then it may make no difference - a bad recording won't necessarily get any worse.

I have heard some recordings which sounded very good, perhaps despite being distributed on MP3 at 160kbps. Some of the issues may not be due to the compression, but the original care in the recording processes.

StanleyB

22-04-2010, 08:06

Why only mention lossy compression. I can clearly detect a difference between FLAC and WAV in the lower notes section.

Tony Moore

22-04-2010, 09:04

Hi Stan,

If you convert from WAV to FLAC and then back again the files match exactly. (I just tried it to make sure I'm not talking rubbish!) So if you can detect a difference when playing FLAC then I can't see how it's the fault of the FLAC format itself, rather of the software or system that you're using to stream it. :scratch:

I've used FLAC and WAV myself on my SqueezeBox based player and I can't tell any difference. What are you using to stream?

Cheers,
Tony

Ali Tait

22-04-2010, 09:11

I guess one advantage is space on your hard drive.I've downloaded some 24/96 albums in FLAC and one album is 1.34 Gb! Wouldn't take too many to fill up the HD.

MartinT

22-04-2010, 09:19

I find mp3 at 256kbps perfectly acceptable for the car and listen all the time while on the move (it helps that my car has a USB socket and shows track/folder info on the main display). I use the good sounding LAME encoder to generate the files.

However, mp3 through a high resolution home system falls way short of acceptable and I never use it there. I like my silver discs, not streaming files.

dave2010

22-04-2010, 10:03

Martin
However, mp3 through a high resolution home system falls way short of acceptable and I never use it there. I like my silver discs, not streaming files.Are you talking about 256 kbps encoded files? I'd still say that you'd be very hard pushed to tell the difference most of the time if the encoding is done properly, but I don't know your system. I'd agree that it can sound awful at rates below that. It also depends on the music. When I listened to some guitar music recently, done at 160 kbps, it sounded very good, but I'd expect full orchestra music to be dire at that rate.

One last thing, some encoders not only do encoding, but also change the dynamics or "normalise" a CD - which I'm sure could wreak havoc. I'm only considering the actual lossy coding to MP3 or AAC, not any other side processes. Also, when I do compression, I always use the highest quality settings for each data rate. Some compressors do a quick and dirty job which they can do quickly even for high bit rates. I have to say I've not actually compared them recently, but my guess is that using low quality encoding could give very audible problems even at high bit rates.

Stan

I'm interested in your comment. Of course the data doesn't get changed if done properly, but that doesn't mean that different streaming systems won't sound different. One issue might be that there are timing problems, or lost data because maybe the processors just can't handle the decompressing in real time. I wouldn't expect that to be the case nowadays though, but a few years back some processors might have struggled with trying to retrieve data from disc and also doing the calculations needed. One other possibility is that there are some issues about the different software players and configurations used to playback the files - as has already been noted in a few threads/posts round here. What software/hardware combinations are you using to play back FLAC vs anything else? Any ideas why you might be getting a different sound?

There aren't too many problems that I know about with the Squeezebox kit, but that tends to have quite sizeable buffering. If you're running a PC or similar maybe you'll get an improvement if you use a larger buffer. It'd be interesting to know.

Ali Tait

22-04-2010, 10:04

Streaming FLAC with the Touch sounds better than any cd player I've tried in my system.

chris@panteg

22-04-2010, 10:45

chrism

22-04-2010, 10:48

I listen to internet radio which is mostly MP3 at 128 kbps and MP3 and Flac from the PC (from CD ripped by me). All is through a wireless modded SB3 and Caiman.

On well recorded stuff I am struggling to tell much difference between the various methods.

I have not heard any hi-res stuff yet though as the SB is 24/44?

Regards

Chris

Peter Galbavy

22-04-2010, 11:16

Depending on what music you listen to you are likely to find that many small scale produced music, especially electronica, is delivered to record labels at 384k MP3s anyway. Certainly how it works for a couple of folk I know. Not their choice, it's the labels.

It probably started out as some expert engineer saying "we need an absolute minimum of 384k MP3" and then some meeja luvvie translated this as "we require 384k MP3". Life's a bitch.

dave2010

22-04-2010, 11:47

Peter
Depending on what music you listen to you are likely to find that many small scale produced music, especially electronica, is delivered to record labels at 384k MP3s anyway. Certainly how it works for a couple of folk I know. Not their choice, it's the labels.

It probably started out as some expert engineer saying "we need an absolute minimum of 384k MP3" and then some meeja luvvie translated this as "we require 384k MP3". Life's a bitch.If I were recording music, and it's not something I do too often, I'd use almost certainly the highest resolution and sampling rate I could lay my hands on which would be compatible with the storage and equipment available. I suppose I might also take into account the final destination, so that if intended for CD I might prefer to record at 44.1 or 88.2 so as to reduce sampling rate change effects, but otherwise I'd tend to use 48 or 96kHz sampling. Any compression would come later in the chain. I might compromise a bit more with speech recording, though it'd probably still be up to decent music standards.

I didn't know that any MP3 systems could even do 384kbps - 320 seems to be the limit for most that I've seen - or are you referring to the sampling rate of the digital input as 384 kHz, which seems more likely to me? Possibly overkill on sampling rate, particuarly if going to be MP3d anyway.

Stratmangler

22-04-2010, 12:35

I didn't know that any MP3 systems could even do 384kbps - 320 seems to be the limit for most that I've seen - or are you referring to the sampling rate of the digital input as 384 kHz, which seems more likely to me? Possibly overkill on sampling rate, particuarly if going to be MP3d anyway.

Dave

It's possible to rip to 348 kbps using dbPoweramp - whether your player will support it is a different matter.
It comes under MP3 Free Format.

dave2010

22-04-2010, 13:53

From the wikipedia entry for MP3 - "Non-standard bit rates up to 640 kbit/s can be achieved with the LAME encoder and the freeformat option, although few MP3 players can play those files. According to the ISO standard, decoders are only required to be able to decode streams up to 320 kbit/s"

There's more detail about standard bit rates and sampling rates in the article, plus quite a lot about the development of the approach. http://en.wikipedia.org/wiki/MP3

Themis

22-04-2010, 17:05

There's a slight "analog" (smoothing) effect when using 256 and 320CBR MP3.
I find that you have much fewer artifacts using 400VBR AAC. It's nearly impossible to distinguish from lossless PCM, except from some very rare "extreme" situations.

If the price of Terrabyte was high, I would seriously consider the option. On portable devices (like iPods), because of various hardware limitations, the 400VBR AAC is indistinguishable from lossless.

dave2010

22-04-2010, 19:32

Dimitri
There's a slight "analog" (smoothing) effect when using 256 and 320CBR MP3.
I find that you have much fewer artifacts using 400VBR AAC. It's nearly impossible to distinguish from lossless PCM, except from some very rare "extreme" situations.

If the price of Terrabyte was high, I would seriously consider the option. On portable devices (like iPods), because of various hardware limitations, the 400VBR AAC is indistinguishable from lossless.Your comments on digitally compressed audio are interesting. If you really can tell the difference between 256 and 320 CBR MP3 and a CD reliably then I suspect that you have (a) better ears than I have and (b) very much better equipment. I'm not going to suggest that you can't though. Similarly I'd expect 400 VBR AAC to be pretty much identical with CD or other high resolution formats for most people, though you have indicated the "extreme" situations where there may be a difference.

My own feelings, with probably lower quality equipment, are that I can't really hear the artefacts for MP3 at 256 kbps, or with AAC at high rates - say 192kbps and above. At lower bit rates the artefacts and sound distortions are unpleasant, at least to me. My own listening suggests that about 95% of the time there can be little difference between compressed audio and CDs, but there are certain parts in music - perhaps only a few seconds in a one hour CD, SACD, DVD or DVD-A, where there is a tingle factor which just doesn't come through with low bit rate compression. I'm just not sure though that higher bit rate compression loses these moments though.

I would also say that generally using better quality equipment does seem to give benefits, even with compressed audio. I've not noticed things sounding worse with better equipment.

I buy SACDs and DVD-As because I do think they sound better, but they may have more information than most of us need theoretically. As the cost of storage is coming down there is really little need to worry about compression from the point of view of disc space. There is a concern about downloading or streaming or broadcasting, but even there, with the exception of wireless broadcasting, it seems likely that there will be internet streaming which could very wel go up to FLAC from CD (and higher) rates in the not too distant future. Where there may be a need for compression for quite a few years is for multi-channel surround sound broadcasting, as the bandwidth requirements for 6 or 8 channels of FLAC equivalent audio is probably going to be too high for perhaps another decade.

Just to put this slightly in context, sitting in the middle of the stalls in the Barbican during a large scale orchestral performance provides a significantly better experience than listening to almost any CD or SACD of similar music. On the other hand, listening in St Pauls Cathedral can seemingly be accurately reproduced by CDs as the acoustic is so muddy that it's easy to reproduce sound which conveys the same impression [e.g Georgiev's recording with the LSO of Mahler's 8th - the performance of which I attended!].

StanleyB

22-04-2010, 19:42

When I was developing the MPress3 (http://www.v3.co.uk/vnunet/hardware/2129160/philex-mpress3-digital-mp3-player) mp3 player in the late 90's, my main worry was the fact that at lower bitrates the higher frequencies are blended together by the mp3 compression process. This very much affected the quality of the sound where the higher frequency are the most important part in the sound.
(I was the 4th person in the world to have developed a mp3 player, and I regret not having made as much money as the iPOD did years later:(. )

Butuz

22-04-2010, 21:27

Well thanks to Stan's DAC I can now tell the difference clearly between 256k MP3 and CD on my low end system. The MP3 sounds muffled in comparison.

I've never heared the difference before but can vouch for it now!

I would hate to imagine how much the difference I hear would be amplified by some of you guys amazing hifi's.... on a 20k system MP3 must sound like fingernails being scratched on a black board!!!

Butuz

Stratmangler

22-04-2010, 21:57

I buy SACDs and DVD-As because I do think they sound better, but they may have more information than most of us need theoretically.

Just bear in mind that with SACD you are buying into an obsolete system of hardware/software. It is a Cul-de-Sac !

At least DVD-Audio is getting a second lease of life now that it is possible to extract the data streams from the discs. And it's this format that has been selected for hi-res downloads.

dave2010

22-04-2010, 23:16

Butuz
Well thanks to Stan's DAC I can now tell the difference clearly between 256k MP3 and CD on my low end system. The MP3 sounds muffled in comparison.

I've never heared the difference before but can vouch for it now!

I would hate to imagine how much the difference I hear would be amplified by some of you guys amazing hifi's.... on a 20k system MP3 must sound like fingernails being scratched on a black board!!!
Butuz
That's very interesting. Maybe I should do some comparison tests then - it's a while since I did any. I never was able to hear much of a difference above 192 kbps before to be honest, though I did feel that a few CDs sounded better at 224 and higher, but I couldn't really say why. Now I've got one of Stan's Dacs it might be time to do a retest.

Chris

Obviously I've taken the cul-de-sac turn already. Some good ideas in SACD, but the determination not to let us get at the bits by the nutters in the music industry has condemned it to death I fear. As you point out, DVD-A now has more life in it. Heh ho - my Oppo SACD/DVD-A player can still sound quite nice, and I'm not too worried about having invested in something which will become obsolete in a few years.

MartinT

23-04-2010, 06:01

Just bear in mind that with SACD you are buying into an obsolete system of hardware/software.

Not quite. The small classical labels continue to support it and new DSD recordings continue to be made, it being fairly trivial these days to convert them to PCM for standard CD release.

Stratmangler

23-04-2010, 09:40

The old Edison Cylinder springs to mind http://en.wikipedia.org/wiki/Phonograph_cylinder - what kind of conversations took place between proponents of the two playback systems available ?

technobear

23-04-2010, 18:36

The opening post was correct. MP3 is the work of the Devil :mental:

I use FLAC now through the Transporter (but I make sure the decoding is done on the PC).

I also have a Sansa Clip+ which handles FLAC and sounds very nice on the move, much better than I've ever heard MP3 sound.

Even at 320kbps, MP3 leaves a lot to be desired. There's so much missing and the sound is tiring to listen to. It's like listening to the audio equivalent of a FAX (remember those?).

Primalsea

24-04-2010, 07:10

We still use the fax machine at work, it one of the most efficient ways of immediate communication there is. You pop your sheet of paper in at one end and it (effectively) appears at the other!

My thoughts on MP3 are that generally its fairly bad but it really depends what music you are listening to. Some music is produced fairly badly to the point that it really doesn't matter how you play it.

Themis

24-04-2010, 07:20

Sure, that Stadium Arcadium can only get better compressed with an MP3 algorithm that what it is on the CD... ;)

Primalsea

24-04-2010, 07:57

Ah, RHCP! Are you a fan too?

Themis

24-04-2010, 08:10

Ah, RHCP! Are you a fan too?
I like the music, but I wished it was better recorded/mixed ! :)

I don't consider myself "a fan", I just have Stadium and One Hot Minute... but I like them anyway. ;)

daveyboy

24-04-2010, 23:22

Mp3's can sound pretty good but they are noticeably flatter than a lossless file which isn't surprising when you consider the compression.

Codifus

25-04-2010, 14:05

Mp3's can sound pretty good but they are noticeably flatter than a lossless file which isn't surprising when you consider the compression.

I think part of the reason that MP3 gets a bad rap is because alot of people do not realize that MP3 quality is determined, significantly, by the encoder.

For example;

in iTunes, the MP3 encdor, even at its best 320 kbps, is lousy. I'm sure the Apple engineers purposely built it that way because "Why do MP3 well when we have AAC? The AAC encoder in iTunes is very good IMO.

LAME is considered one of the best MP3 encoders out there. I've used it to produce some excellent MP3s.

CoolEdit 2K, the digital audio workstation software, also has an excellent MP3 encoder built into it.

Generally, the lousy MP3 encoders, like the iTunes one, are fast. LAME and CoolEdit take quite a while, but the end results are worth it.

CD

dave2010

25-04-2010, 16:08

I think part of the reason that MP3 gets a bad rap is because alot of people do not realize that MP3 quality is determined, significantly, by the encoder.
CDThat is largely my view too. I don't know for sure if the iTunes one is as lousy as you say, but it probably has the defaults set to lowish quality. Maybe it'd be a lot better if all the parameters were tweaked up.

My understanding re MP3 is that the system has been designed so that almost all decoders should give similar results - though I have in the past seen some which claimed to be different/better. I think now most are pretty similar, and should be capable of driving a 24 bit sound card for best results.

On the other hand the encoding process should be adjusted for best results, in which case it will take longer to encode. With a perhaps one off process, or in any case much faster processors these days. There's not too much excuse for not using quality encoding - except perhaps for on the fly real time encoding.

I was actually assuming "high" quality MP3 when I started this post. I think there is a difference, but I find I can't really tell much with some high bit rate encoding. On the other hand there are some really bad encodings, and maybe they're just due to carelessness in the case of some commercially made MP3s or streaming MP3 services. My assumption was that most people round here would use high quality encoding on good originals if wishing to create MP3s or AACs etc.

Codifus

26-04-2010, 11:26

That is largely my view too. I don't know for sure if the iTunes one is as lousy as you say, but it probably has the defaults set to lowish quality. Maybe it'd be a lot better if all the parameters were tweaked up.

iTunes, at its best, 320 kbps, is lousy for making MP3s. It is also very fast. When ever I try an MP3 encoder, I always use it's best setting; 320 kbps CBR. AACs are very good in iTunes, on the other hand. For a short time, I believe it was iTunes 6 Macintosh version, there was actually a LAME MP3 plugin for iTunes. It was great with the iTunes interface for mass encoding a bunch of files at high quality.

My understanding re MP3 is that the system has been designed so that almost all decoders should give similar results - though I have in the past seen some which claimed to be different/better. I think now most are pretty similar, and should be capable of driving a 24 bit sound card for best results.

Call me a cynic, but I disagree. I still think the cheapest encoders are out there all over the place. Fact is, only audiophiles would appreciate a good encoder like LAME, and they are a insignificantly tiny minority when it comes to music. Good encoders are slow, and the last thing that people want on their computer is something that runs slowly. "The best things take time" is rapidly becoming a lost bit of wisdom:)

On the other hand the encoding process should be adjusted for best results, in which case it will take longer to encode. With a perhaps one off process, or in any case much faster processors these days. There's not too much excuse for not using quality encoding - except perhaps for on the fly real time encoding.

I was actually assuming "high" quality MP3 when I started this post. I think there is a difference, but I find I can't really tell much with some high bit rate encoding. On the other hand there are some really bad encodings, and maybe they're just due to carelessness in the case of some commercially made MP3s or streaming MP3 services. My assumption was that most people round here would use high quality encoding on good originals if wishing to create MP3s or AACs etc.

I hope it is now clear that my comments about my experience with MP3 encoders were based on using them at their very best:)

CD

dave2010

26-04-2010, 19:58

David

Thanks for your comments.
iTunes, at its best, 320 kbps, is lousy for making MP3s. It is also very fast. When ever I try an MP3 encoder, I always use it's best setting; 320 kbps CBR. AACs are very good in iTunes, on the other hand. For a short time, I believe it was iTunes 6 Macintosh version, there was actually a LAME MP3 plugin for iTunes. It was great with the iTunes interface for mass encoding a bunch of files at high quality.These points are interesting and it might be worth checking what the options are for the latest iTunes. I use a whole range of different tools, and several use the LAME plugins. I don't know much about AAC encoding - though I do use it. Are there many encoders, or is iTunes the predominant one?

Sadly MP3 is useful, as it works in my car - though I could use WMA instead. I'm guessing that would be worse. It's useful to have quite a lot of things on a few discs, though another approach which I've not really done is to have an iPod plugged into the AUX socket and avoid CDs altoghether, though that's not too helpful while driving. For that application quality is slightly less important, as quite a lot of problems are masked by the noise in the vehicle.

Call me a cynic, but I disagree. I still think the cheapest encoders are out there all over the place. Fact is, only audiophiles would appreciate a good encoder like LAME, and they are a insignificantly tiny minority when it comes to music. Good encoders are slow, and the last thing that people want on their computer is something that runs slowly. "The best things take time" is rapidly becoming a lost bit of wisdom:)

I'm not quite sure what you are disagreeing about here. In my original post I specifically said encoders at this point, not decoders. I do believe that there can be a significant difference between different encoders, or even one encoder with different parameter settings. I'm not sure that I've noticed that encoding is particularly slow even with good encoders and the parameter settings to highest quality, I'm considering reasonably fast modern computers here - though I do have a few older slow ones as well.

I have read that most decoders are substantially the same, though in the past I did think that some sounded different - and some better. Are you saying that you still think that there are some good decoders and some not so good? If so, which do you think are good, and which should we avoid?

Vincent Kars

26-04-2010, 21:20

I think today the differences between encoders are small.
An example of a well conducted listening test: http://listening-tests.hydrogenaudio.org/sebastian/mp3-128-1/index.htm

dave2010

26-04-2010, 23:04

Vincent
I think today the differences between encoders are small.
An example of a well conducted listening test: http://listening-tests.hydrogenaudio.org/sebastian/mp3-128-1/index.htmI think I'm with Codifus/Dave on this one. Re the test you mentioned - "The purpose of this test is to find out which popular MP3 VBR encoder outputs the best quality on bitrates around 128 kbps." - was only at 128 kbps, which I would consider produces pretty lousy results. My own preference would be for at least 192 kbps, and maybe even just set it at 320 CBR. I think I have had good results with VBR between 192 and 320 in the past though - but if there's to be no significant loss of quality, then 320 CBR is usually the one to go for. In that test the iTunes encoder did well, but that does not necessarily mean that it would do best at the higher bit rates. It's possible, but by no means certain. It really does depend on what the purpose of the compression is anyway. Another issue, which I know that some are aware of, is that there can be interference between different compression schemes, so if the source material has already been compressed, then it might be best to use a really high quality compression mode to avoid further problems. An example would be recording from digital radio to MP3 - though of course it'd make much more sense to capture the transport stream if the right tools are available.

If space is no object then don't compress at all, or alternatively use FLAC or ALAC. If compression is desired, but otherwise lossy compress is used, then for compatibility I'd use CBR 320 MP3 at high quality settings, though if I weren't too bothered about compatibility I'd probably use AAC at a reasonably high bit rate.

I think there could be audible differences in the results from different encoders, but the playback should, in theory, sound pretty similar from most decoders.

Themis

27-04-2010, 13:32

If you need encoders' listening test comparisons you have some here : http://soundexpert.org/ratings

StanleyB

27-04-2010, 13:45

I wrote a couple of mp3 ripping software years ago in Lattice C, and remembered that there were a few different DLL libraries floating about that offered different sound qualities. From looking at the .h routines I concluded that the differences were down to how much of the signal was dumped based on the compression ratio and the signal level of the various frequencies.

daveyboy

27-04-2010, 14:06

Well all I can really add is that I have used Easy CD DA extractor, chosen to rip using the slow method and likewise with Express Rip software and still the differences are there. Obviously the lack of body to the sound becomes more apparent the smaller the bitrate, but i can recognise one over the other.

I remember a certain hi fi manufacturer who shall remain nameless was using a link to "prove" there is no audible difference between 128 kbs and lossless, using some kind of easy listening piece. However I was never able to ascertain his point for two reasons, firstly, rock music sounds very muddled when using low bitrates and for that matter the bass on rap music sounds very unlistenable at this bitrate and so the lack of fluctuating transients was obviously going to help. More importantly of course I have usb powered computer speakers which cost all of 7.00 :lol: so i wouldn't appreciate it either way.

Codifus

27-04-2010, 14:51

David

Thanks for your comments.These points are interesting and it might be worth checking what the options are for the latest iTunes. I use a whole range of different tools, and several use the LAME plugins. I don't know much about AAC encoding - though I do use it. Are there many encoders, or is iTunes the predominant one?

I only know of iTunes. Many people try to avoid AAC because it kindof "locks you" into Apple's world with the iPod. If that is indeed the case, is that really so bad? iTunes is a great music server software, and easily one of the most popular. Also, the Ipod is the best selling portable music player out there. The Zune or any other player is a distant 2nd, that distance being comparable to the 193,000,000 miles between the earth and the sun:)

Sadly MP3 is useful, as it works in my car - though I could use WMA instead. I'm guessing that would be worse. It's useful to have quite a lot of things on a few discs, though another approach which I've not really done is to have an iPod plugged into the AUX socket and avoid CDs altoghether, though that's not too helpful while driving. For that application quality is slightly less important, as quite a lot of problems are masked by the noise in the vehicle.

I helped install a high quality iPod adapter into a friend's car. He had a 12 CD changer in the trunk. I showed him the iPod and told him that "this little device is now the equivalent of a 2000 CD changer, so that changer in your trunk has become massively obsolete and should be removed to recover some trunk space." It took him a while to appreciate it. Once he got used to iTunes and the ability to make custom playlists, he was like "OH MY GOD! I can put my entire music collection on the iPod and just drop it in the car!" I'm like, Yeah, that's the idea:)

If you get a high quality ipod car adapter and perhaps the iPod classic with a 160 GB hard drive, you have the makings of many thousands of CD quality songs sitting in your glovebox, ready to play through your car stereo. This brings the phrase road-trip-ready to a whole new level:)

I'm not quite sure what you are disagreeing about here. In my original post I specifically said encoders at this point, not decoders. I do believe that there can be a significant difference between different encoders, or even one encoder with different parameter settings. I'm not sure that I've noticed that encoding is particularly slow even with good encoders and the parameter settings to highest quality, I'm considering reasonably fast modern computers here - though I do have a few older slow ones as well.

I have read that most decoders are substantially the same, though in the past I did think that some sounded different - and some better. Are you saying that you still think that there are some good decoders and some not so good? If so, which do you think are good, and which should we avoid?

You're right. I confused the issue. My apologoes. Encoder quality varies, but decoders I haven't noticed much of a difference. If there is a difference,I beleive it isn't as significant as the variance in encoding quality.

CD

MartinT

27-04-2010, 15:28

Codifus

27-04-2010, 17:19

Even better, order the USB option when you buy a car and you can ditch the i-Pod. Mine takes my 16GB flash drive straight in, shows folders and tracks on the main display and still holds a great many albums which I can control from the steering wheel controls. I use the LAME encoder at 256kbps and it sounds very good.

So, instead of getting an iPod, the better solution would be to buy a new car?
Now why didn't I think of that:lolsign:

I'm sorry, couldn't resist:sofa:

MartinT

27-04-2010, 17:46

So, instead of getting an iPod, the better solution would be to buy a new car?

It had to be done :)

serendipitydawg

27-04-2010, 20:43

How bad is mp3?
Pick your favourite art work. Photocopy it and then photocopy the photocopy.

Then compare the original painting and the last photocopy.

Download from 2L's website some of their free 24/96 (and higher resolution) sampler tracks. One such is recorded in a church. At some point a bird starts chirping and even on my modest equipment it appears to way above the music heightwise.

Make any sort of mp3 from this and I bet you can't even hear the bird, much less its correct height.

Can I sponsor the AOS Creative writing Competition? "MP3: Spawn of the Devil. Discuss"

IMHO Music is very important. Much too important to be compressed to death. Claiming mp3's are musical is verging on disability

The Grand Wazoo

27-04-2010, 23:06

My personal view is that it's a crime to chop up and dispose of large proportions of a piece of art & then to present the result as the equivalent of the original.

However, I also believe that it's possible to enjoy music, despite it being delivered in a manner that is severely compromised.

.........but I'll always prefer the full story better.

MartinT

28-04-2010, 06:14

Nicely put, Chris.

Marco

28-04-2010, 08:54

Indeed.

In a nutshell, MP3 is pish and only fit for undiscerning plebs.

Marco.

Spectral Morn

28-04-2010, 15:54

MP3 is to how the forum worked earlier today and before and full bandwidth music is how it is now....simples.

Regards D S D L

dave2010

29-04-2010, 05:54

Download from 2L's website some of their free 24/96 (and higher resolution) sampler tracks. One such is recorded in a church. At some point a bird starts chirping and even on my modest equipment it appears to way above the music heightwise.

Make any sort of mp3 from this and I bet you can't even hear the bird, much less its correct height.
Do you happen to know which of the 2L tracks has the bird? I'd like to try the test you suggest.

chrism

29-04-2010, 07:11

What a load of "twitter".

Need to get out more and find some real birds!

Regards

dave2010

29-04-2010, 07:17

A couple of fancy ones maybe - http://www.flickr.com/photos/48952723@N00/3779298024/

chrism

29-04-2010, 07:51

Nice legs!

Regards

Marco

29-04-2010, 10:58

Shame about the....

{boat race}

Marco.

serendipitydawg

29-04-2010, 20:01

Do you happen to know which of the 2L tracks has the bird? I'd like to try the test you suggest.

Sorry Dave I don't. I had a brief attempt find "the bird" (Cue juvenile attempts at humour) this morning, without any success

chrism

30-04-2010, 06:59

Shame about the....

{boat race}

Marco.

Don't fancy yours much!

Chris

dave2010

21-05-2010, 09:33

Given a fairly general dislike of MP3 and similar - which I share - all other things being equal - which often they are not, can anyone detect a difference (hopefully positive) between such compressed audio played back through a 16 bit system or through a 20 or 24 bit system?

I currently play tracks from Spotify through my laptop USB to a Caiman, and that limits the resolution to 16 bits. I could perhaps, with seemingly a fair bit of effort, find a way to stream the tracks out to my Squeezebox and then back into the Caiman at 24 bits.

There are other ways I could check out the 24 bit versus 16 bit mp3 (and other) playback. I just wonder if anyone thinks it could make a difference which would be worth having, or if digitally compressed audio is so compromised it's a no hoper.

MartinT

21-05-2010, 09:45

You can't bring back data which is lost. mp3 is hopelessly compromised because it throws away 90% of the data. Changing the bit width from 16 to 24 is going to achieve nothing except to create a more accurate rendition of the flat, lifeless sound :)

Peter Galbavy

21-05-2010, 09:59

I think it's unfair to say 90% - there are actual real numbers somewhere about what MP3 (and other) codecs do, perhaps on the Ogg whitepaper sites. Nevertheless, it does manage to impact the overall delivered sound quite a bit...

MartinT

21-05-2010, 10:46

nb2

21-05-2010, 12:08

Compare the size of extracted WAV files from a CD with, say, 256kbps mp3 files. The latter are usually less than 10% the size of the former, so like it or not it is accurate to say that 90% of the data is discarded.

flac files (generated from those wav files) size are usually about 50% wav files size.
So you are convinced than 50% of the data is discarded when you use flac ...
You have to scream all over the world, that flac files discard audio data.
Because almost everybody is convinced than it does not :scratch:

Peter Galbavy

21-05-2010, 12:34

You might want to meet our old friend Claude Shannon and his new fangled thing called information theory... :)

Codifus

21-05-2010, 13:14

flac files (generated from those wav files) size are usually about 50% wav files size.
So you are convinced than 50% of the data is discarded when you use flac ...
You have to scream all over the world, that flac files discard audio data.
Because almost everybody is convinced than it does not :scratch:

That's a bit of an apples to oranges comparison because mp3 is lossy and flac is not.

CD

MartinT

21-05-2010, 13:19

nb2

21-05-2010, 13:45

I am well aware of Shannon's theory. Even using your compression factor (and yes I do accept that FLAC is lossless), 80% of the original AD coding is lost.

So where does that leave us? The large majority of the original coding data is lost. This includes all the data that encoders such as LAME consider to be 'masked', i.e. inaudible behind other sounds. Our ears do tell us better, however you want to argue it.

Well you accept flac is lossless, so your "proof" using size comparison is not valid.
I really don't know what percentage of data is lost with mp3.
80% seems totally unrealistic to me, can you tell us where this figure comes from ?
This is obviously not from file sizes, otherwise at least 30% to 40% of data would be lost with flac (file size is 50%)
(and I expect this percentage to be very different between 160 and 320 mp3)

Don't get me wrong, I use flac, I don't see any reason to use lossy formats today.

What my ears tell me is that mastering quality is astonishingly more important than mp3 vs flac.

Take a flac file, use lame to generate mp3 at 160, 192 and 320

Then use abx option of foobar to check how easily you can pick up flac from various mp3 qualities.
I'm not saying you won't tell the difference, but quality of mastering is much more important.

And after this test, tell us at which quality you can really pick up mp3 from flac without hesitation, and which tune you used :-)

leo

21-05-2010, 14:31

MP3 music has its place , on a portable MP3 player whilst your on the move , with all the outside noises etc its ok but for serious listening I find it poor

MartinT

21-05-2010, 15:07

Well you accept flac is lossless, so your "proof" using size comparison is not valid.
I really don't know what percentage of data is lost with mp3.
80% seems totally unrealistic to me, can you tell us where this figure comes from ?

Simple maths. WAV (raw encoded PCM) files to mp3 files are approximately 10:1 size ratio. I am using 256kbps files as an example, so not the worst mp3 files by any means.

Accepting that lossless systems like FLAC and MLP can attain 50% packing, that ratio becomes 5:1 lossy ratio.

You could argue that real music files encode to anywhere between 4:1 and 6:1 for good mp3 bitrates. That's still a crazy amount of loss, centering around 80%. No other proof is required, it's there in the numbers.

Try this quote "Uncompressed audio as stored on an audio-CD has a bit rate of 1,411.2 kbit/s,[note 2] so the bitrates 128, 160 and 192 kbit/s represent compression ratios of approximately 11:1, 9:1 and 7:1 respectively." from http://en.wikipedia.org/wiki/MP3

nb2

21-05-2010, 15:38

Simple maths. WAV (raw encoded PCM) files to mp3 files are approximately 10:1 size ratio. I am using 256kbps files as an example, so not the worst mp3 files by any means.

Accepting that lossless systems like FLAC and MLP can attain 50% packing, that ratio becomes 5:1 lossy ratio.

You could argue that real music files encode to anywhere between 4:1 and 6:1 for good mp3 bitrates. That's still a crazy amount of loss, centering around 80%. No other proof is required, it's there in the numbers.

Try this quote "Uncompressed audio as stored on an audio-CD has a bit rate of 1,411.2 kbit/s,[note 2] so the bitrates 128, 160 and 192 kbit/s represent compression ratios of approximately 11:1, 9:1 and 7:1 respectively." from http://en.wikipedia.org/wiki/MP3

The given figures are resulting size ratio figures.
With mp3 we have a complicated mix of data decimation, and lossless size compression, both participating in resulting size ratio.
Those are very different things and you treat them as if they had something in common.
It would be too complicated to give a full overview of compression algorithmic.
Even sticking with lossless compression algorithmic, your reasoning is not valid.
Check flac lossless compression options, you will get different algorithmic speed and different resulting ratio.

I am quite positive that 80% data loss would be awfully unlistenable.

MartinT

21-05-2010, 16:54

nb2

21-05-2010, 17:00

There comes a point where the FLAC or MLP file becomes uncompressible any further without loss because the remaining music data is pseudo-random. That is still a vast amount of data compared with a lossy mp3 file. However the lossy algorithm works, huge amounts of data are discarded!

However, at this point I will duck out. My point is clear for those who want to see it.

I'll just add that your reasoning ignore dynamic range and suppose that all mp3 are cbr, which is plain wrong ...
( check your wikipedia link for cbr vbr and abr, this is the big problem when you just pick up a few lines in an already simplified document dealing about a complex subject )
And I'll duck out there too :)

The Grand Wazoo

21-05-2010, 17:48

Please bear in mind folks, that all this discussion about chucking away data (to create whatever format) is done after the signal has already been digitised - a process whereby a whole lot more information has already been chucked away never to be seen or heard from again.

nb2

21-05-2010, 17:59

MartinT

21-05-2010, 18:26

a process whereby a whole lot more information has already been chucked away never to be seen or heard from again.

Indeed. Were it not so, we would not be able to tell the difference between 16/44 and 24/96 digitisation, for instance.

Themis

21-05-2010, 21:09

I really don't know what percentage of data is lost with mp3.
Well, it's pretty easy to calculate : the compression algorithms are very very few and they all achieve (roughly) the same compression rate. ;)

So, FLAC compression (just like ALAC, OGG or MP3) is the same. So, the ratio between two compressed formats (one lossless and one lossy), gives you the ratio of the lost information.

For instance, AAC400VBR is half the size of a FLAC. Half the information is discarded. Easy. :eyebrows:
AAC256ABR is 1/4 of the size and AAC128ABR is about 1/10th of the size of an ALAC...

Ali Tait

22-05-2010, 07:52

On a subjective level,I find that while cd's I have in FLAC sound great,recordings at 24/96 have much more soundstage depth,and sound more like vinyl than cd,to my ears at least.CD level FLAC can sound very good,but can't quite equal the soundstage depth.This to my ears is what the extra resolution is giving you.It's obvious on my electrostatics,which in general are known for their transparency and imaging abilities.This might not be as obvious on more conventional speakers.

I guess similar will apply to mp3.You could get a good sound,but the soundstage will be even flatter.

nb2

22-05-2010, 09:48

Well, it's pretty easy to calculate : the compression algorithms are very very few and they all achieve (roughly) the same compression rate. ;)

So, FLAC compression (just like ALAC, OGG or MP3) is the same. So, the ratio between two compressed formats (one lossless and one lossy), gives you the ratio of the lost information.

For instance, AAC400VBR is half the size of a FLAC. Half the information is discarded. Easy. :eyebrows:
AAC256ABR is 1/4 of the size and AAC128ABR is about 1/10th of the size of an ALAC...

Well, plain wrong and full of huge misunderstanding ...
part of my job is using, writing, programming compression algorithms for critical applications, and I've been doing that for many years ...
There are tons of algorithms, even as far as audio is concerned.
Your conclusions concerning discarded data using resulting ratio are complete nonsense.
Concerning bit rate, vbr can be implemented in very different ways, more or less efficient.
(private joke : I also have serious background in sound interaction with room if you see what I mean ... :) , and the worst thing is when you are refering, on all forums, to the same apparently serious document, written by somebody who, like you, obviously lacks serious background on the subject )

As I said earlier, this is the big problem when you just pick up a few lines in an already simplified document dealing about a complex subject ...
And in this case, you just have to read the given wikipedia link to check this is complete nonsense.

If people are happy with nonsense, no problem for me :)
After all, it could be worse.
you could try to help cure somebody, spend two hours in a medical library, tell the doctor he doesn't understand anything ... and kill the guy :rolleyes:

I sincerely hope that you NEVER participate in forums where medical or health advice is given ...

Marco

22-05-2010, 10:25

nb2

22-05-2010, 10:33

Just a small point, 'nb2', but what's your first name and where are you from? Did you create a thread in the welcome area when you joined?

If not, please do so at your earliest convenience, supplying the information requested above, and also details of your system and music tastes.

Cheers! :cool:

Marco.

If I am not wrong, you just have to check the list of my posts.
I introduced myself in my very first post on april 29th 2009.
I must confess I gave my first name only the following day ...
you can check that on the post I wrote on 30th april 2009
(this is nicolas from france, paris)

Maximum

22-05-2010, 10:44

I do not know the precise way the mp3 codecs work but it seems unlikely that a direct ratio comparison of mp3 to flac gives anything meaningful in terms of the information preserved. Sure the data will be a lot less, but the information cutting part of the algorithm should make the data compression parts job easier, giving it more redundant data to work with.

nb2

22-05-2010, 10:46

(answering your question, Marco)

My system is almost the same.

Speakers : PMC DB1i

Amp : Creek Evolution 2

I still have my squeezebox classic, but will replace it with a Touch very soon
(will be out for a few weeks, and will buy it when I come back, mid June)

I still have my TC-7520, but I only use it as a usb PC soundcard since I took care of my SB3 power supply.

same musical tastes : almost everything but rap :)

Ali Tait

22-05-2010, 10:59

So could you explain for us mere mortals what information is lost please?

Marco

22-05-2010, 11:04

If I am not wrong, you just have to check the list of my posts.
I introduced myself in my very first post on april 29th 2009.
I must confess I gave my first name only the following day ...
you can check that on the post I wrote on 30th april 2009
(this is nicolas from france, paris)

Thanks for that, Nicolas. I will add those details to your profile :)

Marco.

nb2

22-05-2010, 11:10

So could you explain for us mere mortals what information is lost please?

I'm not sure that it is useful to make a full explanation about that, which is complex, requires scientific background, and is anyway available on Internet.

As far as I am concerned, I just wanted to point out that the percentage of data loss is not fixed.
It highly depends, not only from the algorithm you choose, but also from the options you choose for this algorithm, and even from the data you are compressing
Even with all that well defined, two different softwares can give two different results ...

Concerning lossy audio, you have to mix
1/ discarding data
2/ compressing remaining data

These are 2 very different things.
MartinT and Themis explanations ignore the mere basics of those 2 topics, and are plain wrong.

I just wanted to point out that 80% loss seems totally unrealistic to me.

What is useful and instructive as far as music quality is concerned is to try the test I gave there with foobar abx option :
http://theartofsound.net/forum/showpost.php?p=124300&postcount=61

Themis

22-05-2010, 11:31

Well, plain wrong and full of huge misunderstanding ...
part of my job is using, writing, programming compression algorithms for critical applications, and I've been doing that for many years ...
There are tons of algorithms, even as far as audio is concerned.
Your conclusions concerning discarded data using resulting ratio are complete nonsense.
Concerning bit rate, vbr can be implemented in very different ways, more or less efficient.
(private joke : I also have serious background in sound interaction with room if you see what I mean ... :) , and the worst thing is when you are refering, on all forums, to the same apparently serious document, written by somebody who, like you, obviously lacks serious background on the subject )

As I said earlier, this is the big problem when you just pick up a few lines in an already simplified document dealing about a complex subject ...
And in this case, you just have to read the given wikipedia link to check this is complete nonsense.

If people are happy with nonsense, no problem for me :)
After all, it could be worse.
you could try to help cure somebody, spend two hours in a medical library, tell the doctor he doesn't understand anything ... and kill the guy :rolleyes:

I sincerely hope that you NEVER participate in forums where medical or health advice is given ...
Please keep it objective. You say it's your primary job, but you give no arguments to support that "conclusions concerning discarded data using resulting ratio are complete nonsense". None, nada. So, without precise arguments the "nonsense" is at your side. :lol:

Nobody is impressed by your (self-proclaimed) "knowledge", I assure you. :ner:

DSJR

22-05-2010, 12:03

nb2

22-05-2010, 12:45

Please keep it friendly fellas :)

I have a treasured file at 128Kb/s which sounds amazing to me, although I have to admit that the content isn't too difficult to re-constitute..

Bearing in mind that many "golden ears" have difficulty telling 360Mb/s files from CD Redbook 1441 ones, it makes the whole thing of hi-res files the preserve of audiophools I reckon, but then again, I'm no engineer, let alone scientist and my golden ears have corroded :ner:

Good point :)
Well let's forget about those technical topics.
Some of my close friends are doctors.
They often have to deal with those new internet specialists, who explain to the doctor that he is wrong, and demand explanation or justification for 10 years studying knowledge, when themselves don't have 5% of the necessary backgroung to understand ... :lol: As long as their health is not in danger, the doctor says "yes you are right", and let them take a totally useless medicine, totally unrelated to their problem ... :rolleyes:

The real subject being mp3 audio quality, and not the relation between file size and quality, try to make the test I talked about.
It is very easy and you learn much more than with tedious technical and theoretical blabla.
I do it regularly and I am still often surprised by the result.
My findings so far is that good mastering is the most important point, by far.

This actual experiment, leads to facts, which are contradictory with beautiful theoretical explanations of some folks :)

Marco

22-05-2010, 13:27

Please keep it friendly fellas :)

Absolutely. I would echo Dave's sentiments; indeed I insist on them! ;)

So let's respect everyone's point of view, even if we don't necessarily agree with it :cool:

Marco.

nb2

22-05-2010, 13:56

I realized I wasn't very clear for the foobar abx test

In foobar 2000 v1.0.3
- select exactly two tracks in a playlist (namely one song in flac, and the same song in mp3 generated from this flac)
- right click on them
- use "Utilities" -> "ABX two tracks ..."

Have fun and give us the results :)

PS :
All mp3's must be generated from the flac.
For obvious reasons, you are supposed to get a better quality mp3 160k with
flac -> mp3 160k
than with
flac -> mp3 320k -> mp3 160k

Peter Galbavy

22-05-2010, 18:25

I have to agree with Nicholas on the whole thing but perhaps he is more lucid and can be bothered arguing. As someone who has written image compression software (JPEG and proprietary) and read quite a bit on audio and video compression it seems to me that a number of the "subjective" crowd here are using sheer bluster, seniority and hand waving arguements to try to make a point. There is a wealth of literature out there in both printed and online form that talk about what *real* engineers and scientists do in the field and not "the file is only 10% the size so you must have thrown away 90% of the data".

The point of compression (in general) is to reduce resource consumption (storage, transmission bandwidth etc.) by removing duplicate or redundant data. Lossy compression goes one step further and reduces perceived redundacy by applying design choices to the original data based on experimentation, research and levels of loss that are deemed acceptable (the bits per second stuff). Lossy compression of different types of data each use different criteria to select what is considered OK to ignore - it's worth noting, for example, that even cartoons vs real film for moving images are treated very differently. In the audio world the same choice are made between music, speech, collapsing multiple channels together etc.

Some background that takes 2 mins of google:

http://xiph.org/vorbis/doc/vorbis-fidelity.html
http://en.wikipedia.org/wiki/Psychoacoustics
http://en.wikipedia.org/wiki/Audio_compression_%28data%29

Themis

22-05-2010, 19:14

Peter, whether (or, rather, how) the data removed are perceived is a different subject altogether.

The data is shrinked by an approximate 90% ratio on 128kCBR. The research and experimentations aim to minimize the perceived effects.

But, what is indisputable is that 90% of the information is lost.
Of that, there's nothing to discuss about, I suppose. ;)

Beechwoods

22-05-2010, 19:26

Marco

22-05-2010, 19:33

I'll be writing a thread soon, describing my experiences when I heard a superb system containing (amongst other things) a £12000 Linn Klimax DS, streaming high-quality flacs, and a classic Kenwood L-07D direct-drive T/T with a Linn Troika, both playing the same music - and how the Kenwood comprehensively annihilated the Klimax in every imaginable sense.........

Never mind total shit, piss-poor, MP3!!! :rolleyes:

Look out for it early next week - it's sure to be a thought-provoking read ;)

Marco.

Themis

22-05-2010, 19:40

Remember that a proportion of the compression applied to MP3's and other Lossy formats is actually 'non-lossy data compression' like Winzip would do - not all the 90% reduction in filesize is due to lost data - a lot is - but some is non-lossy and retrieved bit-accurately on decoding. Martin alluded to this earlier.
Nick, I beg to disagree. The bitrate (128kbits/s) is uncompressed. You can't stream more (or less, for that matter) data than the one stored on the media.

A Redbook is about ten times this bitrate. So, there's ten times less information to stream...

No ? ;)

DSJR

22-05-2010, 20:31

Beechwoods

22-05-2010, 20:32

Dimitri - there is a non-lossy compression step built into MP3. As a reference check this link to a chapter from an O'Reilly book on MP3: http://oreilly.com/catalog/mp3/chapter/ch02.html#71109, the MPEG Audio Compression in a Nutshell section:

... The person doing the encoding can specify how many bits should be allotted to storing each second of music, which in effect sets a " tolerance" level-the lower the data storage allotment, the more data will be discarded, and the worse the resulting music will sound. The process is actually quite a bit more complex than that, and we'll go into more detail later on. This kind of compression is called lossy, because data is lost in the process. However, a second compression run is also made, which shrinks the remaining data even more via more traditional means (similar to the familiar "zip" compression process).

My point is only that not all the compression achieved within the MP3 codec is lossy. It's only a fine one though. I don't think it detracts from the main point that MP3 compression denies the listener the opportunity to experience the music as it was originally recorded. From a hi-fi perspective it's a major compromise, subjectivist or objectivist - it ain't as good as non-lossy let alone good analogue!

Marco

22-05-2010, 20:43

From a hi-fi perspective it's a major compromise, subjectivist or objectivist - it ain't as good as non-lossy let alone good analogue!

Indeed! Only those satisfied with grossly sub-standard reproduction of their favourite music would tolerate its significant sonic deficiencies.

Count me (most firmly) out.....!

I would expect most people with discerning ears on AOS to similarly reject it.

Marco.

nb2

22-05-2010, 21:12

The image file formats analogy is very good.
Visual quality is less subjective and easier to share with other people than audio quality.

Like in audio, you have a lot of compression algorithms, some lossless, some lossy.

It is easy to take a picture of very good quality, and convert it to different file formats (png, bmp, jpg, tiff, etc ...) with different options.
It is very easy to check that looking only at size ratios tells you almost nothing about resulting quality.

The same conclusion applies to audio.

All in all, actual experiment is in the end the best test.
I wouldn't use lossy mp3, since lossless flac is available.
But I'd like some of you to try the foobar abx test I explained, and tell us the result.
(try also with mp3 vbr, and check the difference with mp3 cbr )

Themis

22-05-2010, 21:25

Dimitri - there is a non-lossy compression step built into MP3. As a reference check this link to a chapter from an O'Reilly book on MP3: http://oreilly.com/catalog/mp3/chapter/ch02.html#71109, the MPEG Audio Compression in a Nutshell section:

My point is only that not all the compression achieved within the MP3 codec is lossy. It's only a fine one though. I don't think it detracts from the main point that MP3 compression denies the listener the opportunity to experience the music as it was originally recorded. From a hi-fi perspective it's a major compromise, subjectivist or objectivist - it ain't as good as non-lossy let alone good analogue!
Nick, that's exactly what I was saying. There are two compressions involved: An audio compression (lossy) and a algorithmic (lossless).
A lossless compressed file has only algorithmic compression, while a lossy codec has both of them.

Algorithmic compression has the same efficiency on any file stored. Be it lossless or lossy.

That's why the bitrate ratio between lossless and lossy is the same as the file size ratio between the same two files, which is the same as the amount of information discarded : algorithmic compression is the same on both of them. ;)

Please note, that in my file size comparisons I always compared compressed lossless files (FLAC, ALAC, WMA lossless) to lossy compressed files (MPx). As the algorithmic compression is the same, the expanded/compressed ratio should be roughly the same.

MartinT

22-05-2010, 21:34

it seems to me that a number of the "subjective" crowd here are using sheer bluster, seniority and hand waving arguements to try to make a point. There is a wealth of literature out there in both printed and online form that talk about what *real* engineers and scientists do in the field and not "the file is only 10% the size so you must have thrown away 90% of the data".

Just to correct you on your pointed remark, I am an engineer.

nb2

22-05-2010, 21:44

Well I need a break for joke ... :):):)
One day, some guy will very seriously tell you that when you drop an apple, it flies to the sky.
Don't react, otherwise here is what will happen :

1/ If you beg to disagree, he will find scientific explanation (with useful internet links) to prove that dropped apples fly to the sky.
2/ If you tell him to actually try to drop an apple, he won't even notice. It is useless, because with its own personal understanding of Newton theory, he can prove that apples do fly to the sky.
3/ If you tell him that he may have missed some points in Newton theory, he will say no, and ask you to prove your version of Newton theory (and please stop trying to impress people with your self-claimed knowledge ...).
4/ If you are crazy enough to try to explain Newton theory, he will tell you that Newton theory is wrong, and that Einstein theory obviously shows that apples do fly to the sky (new funny internet links about Einstein theory appear at this moment).
5/ Normally at this point, even if you tried to explain Newton theory, you give up theoretical discussions, leave Einstein rest in peace, and note that you should find some time to meditate about human brain relativity ...
6/ Of course, he still doesn't want to check and drop an apple, but, if you have the opportunity to force him, he will throw the apple to the sky, run before the apple actually falls down, and tell you he is right ...
7/ If you are still not insane, and drop 100 kgs of apples on his head, he will forget everything, get back to point 1/ and the story restart over and over ...

Beware, the guy is dangerous for your mind :eek::eek:
Break done ... :):):):cool::cool::cool:

Now, please, a volunteer to explain what vbr is ?

Stratmangler

22-05-2010, 21:47

Now, please, a volunteer to explain what vbr is ?

:D

Stratmangler

22-05-2010, 21:51

My apologies for laughing Nicolas, but you've made a real issue about your knowledge of MP3 algorithms and you ask what VBR is.

It stands for Variable Bit Rate.

nb2

22-05-2010, 21:53

My apologies for laughing Nicolas, but you've made a real issue about your knowledge of MP3 algorithms and you ask what VBR is.

It stands for Variable Bit Rate.

I know ... :lol::lol:
I want somebody to explain for others, because me, I give up :cool: :cool:

nb2

22-05-2010, 21:59

And if you can make them understand, that their reasoning about data loss using bit rate explanation, doesn't fit with vbr, you deserve a gold medal ... :cool:

Stratmangler

22-05-2010, 22:29

Weeeellllllll........

I have a library of FLAC files.
This library is losslessly compressed, and can be returned to its' pre compression state if so desired. To do so will produce identical WAV files to the ones you had prior to applying the lossless compression.

I do not like the idea of MP3 files.
Irrespective of how the things sound, there is a big chunk of data thrown way in the compression process.
Once gone, this data is gone forever.
You cannot accurately rebuild the data to its' pre compression status.
End of argument.

Themis

23-05-2010, 07:06

Irrespective of how the things sound, there is a big chunk of data thrown way in the compression process.
Are you sure ? :scratch:
And perhaps the bigger the audio compression is, the bigger the chunk is thrown away ? :scratch:
Would you even go up to saying that the bigger the chunk thrown away the smaller the file size is ? :eek:

Are you saying that apples "fly to the skies" ? :lol:

Stratmangler

23-05-2010, 08:31

Are you sure ? :scratch:
And perhaps the bigger the audio compression is, the bigger the chunk is thrown away ? :scratch:
Would you even go up to saying that the bigger the chunk thrown away the smaller the file size is ? :eek:

Are you saying that apples "fly to the skies" ? :lol:

Yup, them flying apples are to blame.;)

nb2

23-05-2010, 09:26

Irrespective of how the things sound, there is a big chunk of data thrown way in the compression process.

The question is how big ?
You told us that you know that vbr stands for Variable Bit Rate, congratulations
:cool:
Now have you understood what it means, what is the difference with cbr, and why conclusions about quality vs file sizes are not valid ??? :)
Next volunteer ... :acid:

Themis

23-05-2010, 09:34

The question is how big ?
If you need a number, it's 42 ! :lol:

Another way of evaluating VBR is to listen to it: a AAC128ABR is the grossly the quality of AAC100VBR. And grossly the same size, too. ;)

nb2

23-05-2010, 09:39

Ok we see that you can throw an apple very high.
But don't run so fast, you will hurt yourself :lol:
Someone else can answer the question about mp3 vbr (forget aac, we were talking about mp3) ? :cool:

Themis

23-05-2010, 09:44

Ok we see that you can throw an apple very high.
But don't run so fast, you will hurt yourself :lol:

Pardon me ?

I don't understand, what do you mean ? :scratch:
Listening is not a valid protocol, perhaps ? ;)

nb2

23-05-2010, 09:48

I don't understand, what do you mean ? :scratch:

Are you sure ? I had some doubts about that ...
And no, listening to AAC is not a valid protocol to evaluate mp3 vbr quality ... :lolsign:

Themis

23-05-2010, 09:54

Are you sure ? I had some doubts about that ...
And no, listening to AAC is not a valid protocol to evaluate mp3 vbr quality ... :lolsign:
Stop insinuating and double-meaning posting nb. You're being rude. We are no friends, so please keep it clear and meaningful.

What can be applied to AAC can be applied to MP3, and the listening to MP3VBR vs MP3ABR yields (to no surprise) the same results. So ? :eyebrows:

nb2

23-05-2010, 10:06

Keep it cool.
Are you saying rude ?
I am not trying to impress anybody with my self-claimed knowledge, as you insinuated. ;)

Please someone else explain to him that the question was evaluating quality from file size ratio, and that vbr goes against that.
Now we reach step 4/ of throwing apples.:)

Well time for lunch and back to life ... :cool:
See you later :)

Themis

23-05-2010, 10:20

Please someone else explain to him that the question was evaluating quality from file size ratio, and that vbr goes against that.

Well, when you compare two VBR (a FLAC-950VBR and a AAC-100VBR) then the file size does help to know the amount of information chunk out. ;)

And the same applies when you compare two ABRs (or whatever).

And if you *really* need to compare two different bitrate methods, then converting one to another and listening will help compare what is comparable.
As I said, a 128ABR sounds the same as a 100VBR. And (surprise!!!) the file sizes are roughly the same.

So ? Where's your argumentation (apart from the usual double-meaning "humor") ?

Stratmangler

23-05-2010, 12:19

The question is how big ?
You told us that you know that vbr stands for Variable Bit Rate, congratulations
:cool:
Now have you understood what it means, what is the difference with cbr, and why conclusions about quality vs file sizes are not valid ??? :)
Next volunteer ... :acid:

Which part of Irrespective of how the things sound, there is a big chunk of data thrown way in the compression process are you having trouble understanding ?

Personally, I'd rather retain the data, hence the use of FLAC.

If you drop the condescending attitude you'd get along better with folks.

StanleyB

23-05-2010, 13:35

Personally, I'd rather retain the data, hence the use of FLAC.
And that's why I prefer WAV over FLAC. I can hear a loss of detail in the bass in a FLAC file.

Rare Bird

23-05-2010, 13:37

nb2

23-05-2010, 13:52

Which part of Irrespective of how the things sound, there is a big chunk of data thrown way in the compression process are you having trouble understanding ?

Personally, I'd rather retain the data, hence the use of FLAC.

If you drop the condescending attitude you'd get along better with folks.

I have no problem understanding it, I am saying it is wrong.
First of all let's be clear, I use flac, because I am sure that no CD audio data is lost and I like it.
I also know that mp3 "can" loose data especially mp3 cbr (but I think it is usually a very small loss if you use it carefully and stick to vbr).
So, like you do, I also avoid mp3.

Now why is your statement concerning mp3 vbr plain wrong ?

- because it says big chunk of data is lost (I disagree, but let's say it is true, we will save time)
- because you implicitly claims that big chunk of AUDIO data is lost.
- because you implicitly claims that it results in big difference in audio quality

I am saying that especially for vbr (and all erroneous claims I read here, implicitely take for granted that cbr is used)
- the data lost can be very small and even null.
- the audio data lost can be very small, and even be null.
Yes I insist that data and audio data are different things.

Why data loss can be small and even null with vbr ?
This is explained all over internet, check Wikipedia.
This is the main goal of vbr concept.
You tell me I am condescendant ?
I find it condescendant to deny it without even checking, this is the purpose of vbr, and you deny it ...

I insist on this difference because the marketing bullshit saying that audio quality with 192 Khz is better than with 88 Khz relies on this lie.
To be clearer, let's give two examples.
1/ tags in a flac are data, but are not audio data, no effect on audio quality ...
this is obvious I hope so ...
2/ in a flac coding PCM data at 192 Khz, the highest frequencies you can possibly store are not audio data.
So if you downgrade it to a 88 Khz flac, you may loose data, but you won't loose any audio data.
This is less intuitive, you have to understand Nyquist theorem and signal theory to understand why NO AUDIO DATA IS LOST and why AUDIO QUALITY IS EXACTLY THE SAME.

The subject of the thread is mp3 audio quality, I am interested in that.
If some people are convinced that Nyquest theorem is false, and that all people working with signal theory for decades are morons, after all I don't care.
I use the foobar ABX test I explained, which is the best test I found as far as mp3 audio quality is concerned.
If you have a better one, I'll be happy to try it.

Now, you still claims I am wrong.
Honestly I would prefer that you drop your condescendant attitude and try the test, and discuss the results, than accusing me of being condescendant.

If some people are interested in mp3 audio quality with actual experiment with my test, or any other test we could find better than this one, and discuss it, I will be happy to participate.
Now as far as pseudo scientific blabla is used to lead to wrong conclusions, if someone else want to discuss it with an open mind, please contact me by private message.
Here I think we are in a dead end.

nb2

23-05-2010, 13:54

I've a great test for you, copy a live (Concert) CD to MP-3 & notice the micro pause between tracks (no tracks as such on a live album as it's suppose to be a continuation, but the CD has to tell the CD Player it's moving tracks, well MP-3 pick up these in the form of a pause).

I guess this comes from your software extraction options rather than from file format.
This is well documented in Exact Audio Copy documentation.

Rare Bird

23-05-2010, 13:59

You get one every day :rolleyes:

nb2

23-05-2010, 14:05

You get one every day :rolleyes:

Have you tried optimal gapless playback with foobar ?
http://wiki.hydrogenaudio.org/index.php?title=Gapless
http://www.foobar2000.org/

Rare Bird

23-05-2010, 14:29

What makes you think i will entertain MP-3. :lolsign:

nb2

23-05-2010, 14:32

What makes you think i will entertain MP-3. :lolsign:

You told me that you tested with mp3 ...
I don't use it either, except for my quality tests. :lolsign:

Rare Bird

23-05-2010, 14:58

You told me that you tested with mp3 ...

No i didnt i said **heres a test**

Themis

23-05-2010, 15:09

Now why is your statement concerning mp3 vbr plain wrong ?

- because it says big chunk of data is lost (I disagree, but let's say it is true, we will save time)
- because you implicitly claims that big chunk of AUDIO data is lost.
- because you implicitly claims that it results in big difference in audio quality

I am saying that especially for vbr (and all erroneous claims I read here, implicitely take for granted that cbr is used)
- the data lost can be very small and even null.
- the audio data lost can be very small, and even be null.
Yes I insist that data and audio data are different things.

Why data loss can be small and even null with vbr ?
This is explained all over internet, check Wikipedia.
This is the main goal of vbr concept.
You tell me I am condescendant ?
I find it condescendant to deny it without even checking, this is the purpose of vbr, and you deny it ...

I insist on this difference because the marketing bullshit saying that audio quality with 192 Khz is better than with 88 Khz relies on this lie.
To be clearer, let's give two examples.
1/ tags in a flac are data, but are not audio data, no effect on audio quality ...
this is obvious I hope so ...
2/ in a flac coding PCM data at 192 Khz, the highest frequencies you can possibly store are not audio data.
So if you downgrade it to a 88 Khz flac, you may loose data, but you won't loose any audio data.
This is less intuitive, you have to understand Nyquist theorem and signal theory to understand why NO AUDIO DATA IS LOST and why AUDIO QUALITY IS EXACTLY THE SAME.

The subject of the thread is mp3 audio quality, I am interested in that.
If some people are convinced that Nyquest theorem is false, and that all people working with signal theory for decades are morons, after all I don't care.
I use the foobar ABX test I explained, which is the best test I found as far as mp3 audio quality is concerned.
If you have a better one, I'll be happy to try it.

Now, you still claims I am wrong.
Honestly I would prefer that you drop your condescendant attitude and try the test, and discuss the results, than accusing me of being condescendant.

If some people are interested in mp3 audio quality with actual experiment with my test, or any other test we could find better than this one, and discuss it, I will be happy to participate.
Now as far as pseudo scientific blabla is used to lead to wrong conclusions, if someone else want to discuss it with an open mind, please contact me by private message.
Here I think we are in a dead end.
Ok, enough of this theories about VBR. Visibly you talk about what the bitrate target can do, and the (supposed) no-link-with-file-size, but you miss some concrete data. :)

I can provide you some figures, if you wish :

I have a library of 7500 songs. They are in two formats: Both in FLAC and AAC400VBR.

The average VBR of FLAC in my library is : 858.3kbps
The average VBR of the AAC400 is : 353.7 (because you probably don't know or you omit on purpose to mention that the VBR target rule of MPx codecs does not produce the exact bitrate -400- but an best-effort approximation)
So, the ratio between the bitrates is : 2.42

The total size of the FLAC library is : 858.3MB
And the AAC400VBR one is : 79MB
The ratio is : 2.38 (surprise!!! roughly the same as the bitrate ratio... :eyebrows:)

Now, one can object that in individual songs the ratio may be different (remember ? the VBR misunderstanding ? ;))
Let's take four songs : the first, the last and a couple in the middle (10cc, Madness, Tchaikovsky and ZZ Top).
The VBR rates (kbps) and ratios and file sizes (in GB) and ratios are:
1. 933 / 378 (ratio 2.46) 19.6 / 8.1 (ratio 2.41)
2. 668 / 298 (2.24) 5.3 / 2.4 (2.2)
3. 480 / 314 (1.52) 10.5 / 6.9 (1.52)
4. 828 / 324 (2.55) 24.2 / 9.5 (2.54)

Conclusion : the AAC400VBR codec reduces always the bitrate and filesize consistently. I wonder how even one can be naive enough to think the contrary. :mental:

Now, back to audio quality:

The loss of data in audio quality for lossy codecs follows a smooth line up to 100VBR (or 128CBR), which means that the subjective loss is smaller than the amount of data discarded up to that point.
Which doesn't mean that there is no loss (and still less that there is no data discarded). Simply, by discarding 50% of the data you discard a much lesser quality percentage (say 3% to give an approximation) : this is the case of AAC400 which is roughly half the size of FLAC. AAC100VBR discards about 20% of the subjective quality (my estimation) but manages to lower by 85% disk space.

From 100VBR and under, things get tougher, and even under 48kbps the situation is inversed : codecs start discarding huge amounts of subjective quality compared to the disk space saved. That's why these compressions are not really used for music.

I guess this closes the pseudo-subject of bitrate/filesize hype. The consistency is proved by the figures. F*ck the theory. :)
And please, keep ABX tests away from this. They have proven being wrong so many times that they are completely useless. I can provide you with some consistent statistical data (not ABX bullshit), if you need to compare audio codecs.
Or better : use your own ears ! :)

Any objections ? :cool:

PS: Can you please explain to me why the heck are you talking about the Nyquist–Shannon theorem ? What does it have to do with the subject ? :scratch:

Rare Bird

23-05-2010, 15:38

:popcorn:

Maximum

23-05-2010, 16:18

If you encoded a file to 400 then surely the ratios will be roughly the same by default, as that is the target you have set for the encoder? Surely that doesn't mean that the quality loss is necessarily consistent with all the files? Less complex data should be easier to compress. A file of silence would be very easy to compress with compression designed to take advantage of it, as only a single sample would need to be stored with it's multiple, even though this is obviously a rather extreme example.

This is getting quite heated, I hope we can all keep civil about this.

Themis

23-05-2010, 16:25

If you encoded a file to 400 then surely the ratios will be roughly the same by default, as that is the target you have set for the encoder?
The target for VBR encoding I use is is the one by default

Surely that doesn't mean that the quality loss is necessarily consistent with all the files?
Well, interesting, indeed.

There are two ways of measuring the target quality:
1. The one of the encoder: the encoder has a constant quality target (I don't change settings between albums, let alone songs). So, unless the encoder is buggy (that I doubt) the quality target should be consistent.
2. My ears : I have to say that I noticed no inconsistencies, although I listen to my library almost the all day through. I'm fussy, so I should have noticed if there were differences in quality between tracks, but I didn't.

So, I suppose the risk that such an inconsistency exists is *really* minimal, imho.

Peter Galbavy

23-05-2010, 16:41

And that's why I prefer WAV over FLAC. I can hear a loss of detail in the bass in a FLAC file.

Then, Stan, your player is broken. They should result in the same thing. It's like saying you prefer keeping your letters as .doc files because then you ZIP them they come out different...

Peter Galbavy

23-05-2010, 17:28

Beechwoods

23-05-2010, 17:45

The simple test is to listen to how it sounds. At the end of the day I couldn't care less about the theory if the end result was the music sounded worse than when it started out.

The only lost data from a WAV > FLAC > WAV transcode will pertain to WAV header info not supported by FLAC. The difference Stan hears is more than likely down to the reliability / accuracy of the realtime FLAC decoder, rather than a broken CD player ;)

Primalsea

23-05-2010, 18:55

I must admit I have been following this and I am now lost on a subject that I have never really thought about and can barely be bothered to think about it now!

It seems to me that this "discussion" isn't actually about the respective points that both side are making. It seems to be one about nomenclature and what it means differently to both parties.

It seems true to me that if you compare a file to its compressed counterpart then effectively you have lost data. If the counterpart has less bytes than the original then data is lost as bytes = data does it not?

However its not so much the data that is important, its the end result obtained from putting that data to use, ie. listening to music.

So you can look at it another way and say that the data per se has not been lost just encoded more efficiently so that effectively the same data is retained just stored in a different way so that less actual bytes are required while still being able to reproduce the same end result as the original file.

The how much of the data in the original file was of any use anyway question is just another dynamic which just complicates things. Data relating to out of band noise that that is below the noise floor is just the start. It continues to expand until you then get into arguments about what is perceivable and what's not.

At this point you often get people trying to tell you what you can and can't hear rather that asking you.

dave2010

23-05-2010, 22:54

Paul
I must admit I have been following this and I am now lost on a subject that I have never really thought about and can barely be bothered to think about it now!

It seems true to me that if you compare a file to its compressed counterpart then effectively you have lost data. If the counterpart has less bytes than the original then data is lost as bytes = data does it not?

However its not so much the data that is important, its the end result obtained from putting that data to use, ie. listening to music.

So you can look at it another way and say that the data per se has not been lost just encoded more efficiently so that effectively the same data is retained just stored in a different way so that less actual bytes are required while still being able to reproduce the same end result as the original file.

The how much of the data in the original file was of any use anyway question is just another dynamic which just complicates things. Data relating to out of band noise that that is below the noise floor is just the start. It continues to expand until you then get into arguments about what is perceivable and what's not.

At this point you often get people trying to tell you what you can and can't hear rather that asking you.Quite a number of folks around here seem to be getting muddled.

Lossless compression means just that. There is no loss of information. The data may be lost in the sense that files become smaller, or streams of data require lower bandwidth, but the original data can be reconstructed exactly. On the other hand, compression such as MP3 is lossy, and not only is there what you might call a loss of data, but there is also a loss of information. Thi information can never be recovered precisely.

We are talking about data from a digital source which has already been encoded. If the data is originally from an analogue source there may be some loss of information in converting it to digital. This is a separate issue.

Lossy compression can be very effective if priorities are for small file sizes and/or low bit rates suitable for streaming. There is a loss of quality, and this tends to be more severe if the compression is hard. Conversely, for miilder compression, the quality loss is not always considered unreasonable. Most lossy compression methods with bit rates of 256kbps or higher would generally be considered acceptable, though not by those who have very high quality equipment, or who are very fussy about what they hear. Most objective testing has shown that most people are not able to detect the difference between a certain level of lossy compressed file and the original by listening, though often the tests are carried out on short sections of music, and the music may not be representative. Look at it this way, there may be just a couple of seconds in an hour long piece which send tingles down your spine if done well. Any comparison test will probably not set out to test on such passages of music, and thus it will tend to confirm that most of us can't tell the difference most of the time. It's at those crucial moments that one probably can tell a difference, but that's not what most of the more objective approaches to evaluation actually do.

I have highlighted in blue in your orignal text part of a sentence which applies well to lossless data compression, but which is untrue for lossy compression.

StanleyB

24-05-2010, 00:08

The only lost data from a WAV > FLAC > WAV transcode will pertain to WAV header info not supported by FLAC. The difference Stan hears is more than likely down to the reliability / accuracy of the realtime FLAC decoder, rather than a broken CD player
And it's there where I have an issue with the assumption that FLAC is lossless. My reasoning is that if the encoder/decoder produces a difference between the original WAV and the decoded FLAC to WAV, then FLAC is not entirely lossless. The instant that an encoder/decoder produces a difference, you are experiencing a loss, and thereby turning your lossless file into a lossy one.

Codifus

24-05-2010, 00:31

Themis

24-05-2010, 05:36

And it's there where I have an issue with the assumption that FLAC is lossless. My reasoning is that if the encoder/decoder produces a difference between the original WAV and the decoded FLAC to WAV, then FLAC is not entirely lossless. The instant that an encoder/decoder produces a difference, you are experiencing a loss, and thereby turning your lossless file into a lossy one.
Well, I've made several tests (and some other people, too) and the problem when there are differences don't come from the FLAC decoder. The decoder is perfect. And very simple.

No, the difference (when there is one) comes from the fact that the stream decoder is done by a dedicated chip which *may* interfere in several ways with the rest of the audio circuitry. When there is a difference, it is simply because of the mediocre design of the streamer at various other levels.

You can do a simple experiment:
Stream a WAV, then the same WAV converted to FLAC then back to WAV. If you hear differences between the two (identical) WAVs then you have a problem.
Same applies for ALAC , or every other lossless format. ;)

Also, let's not forget that all studio hi-rez material is sold (online) as FLACs, ALACs or WMA lossless. If there was a possible difference, nobody would use these formats. ;)

Themis

24-05-2010, 05:46

Lossy compression can be very effective if priorities are for small file sizes and/or low bit rates suitable for streaming. There is a loss of quality, and this tends to be more severe if the compression is hard. Conversely, for miilder compression, the quality loss is not always considered unreasonable. Most lossy compression methods with bit rates of 256kbps or higher would generally be considered acceptable, though not by those who have very high quality equipment, or who are very fussy about what they hear. Most objective testing has shown that most people are not able to detect the difference between a certain level of lossy compressed file and the original by listening, though often the tests are carried out on short sections of music, and the music may not be representative. Look at it this way, there may be just a couple of seconds in an hour long piece which send tingles down your spine if done well. Any comparison test will probably not set out to test on such passages of music, and thus it will tend to confirm that most of us can't tell the difference most of the time. It's at those crucial moments that one probably can tell a difference, but that's not what most of the more objective approaches to evaluation actually do.

To my experience, lossy compression acts a bit like a tape recorder :
- A low-level noise is mixed with the music (just like tape hiss), then
- All information under this noise is stripped. Not a real problem, as it's below the noise.

The bigger the compression rate, the "cheaper" the tape recorder. ;)

There's a loss in SNR and in low-level detail. But it is not crucial for several everyday applications (car radios, portable devices with mediocre built-in speakers or earbuds)

MartinT

24-05-2010, 05:52

The bigger the compression rate, the "cheaper" the tape recorder. ;)

Nice analogy, except that I find the effects of cheap digital are far harder to listen to than cheap analogue. For instance, listening to an old cassette gives a kind of 'radio' rendition which at least preserves the general thrust of the music while rolling off the bandwidth extremes. Listening to DAB radio, for instance, on a poor broadcast is much, much worse with horrible digital artifacts across the spectrum.

Themis

24-05-2010, 05:55

Nice analogy, except that I find the effects of cheap digital are far harder to listen to than cheap analogue. For instance, listening to an old cassette gives a kind of 'radio' rendition which at least preserves the general thrust of the music while rolling off the bandwidth extremes. Listening to DAB radio, for instance, on a poor broadcast is much, much worse with horrible digital artifacts across the spectrum.
I have to agree. Unfortunately compression under 96k introduces some very unpleasant distortion. :)
I'm probably too old.

StanleyB

24-05-2010, 06:26

You can do a simple experiment:
Stream a WAV, then the same WAV converted to FLAC then back to WAV. If you hear differences between the two (identical) WAVs then you have a problem.
Been there, done that and more, even got the CD.

Also, let's not forget that all studio hi-rez material is sold (online) as FLACs, ALACs or WMA lossless. If there was a possible difference, nobody would use these formats. ;)
Very weak assumption. Since when did the studios start showing us that much respect?

Themis

24-05-2010, 06:29

Been there, done that and more, even got the CD.
And ? Did you find a difference between the two WAVs ? :scratch:

(The CD is another problem altogether)

Very weak assumption. Since when did the studios start showing us that much respect?:lol:
Yeah, you're probably right. ;)

Primalsea

24-05-2010, 06:43

The digital data is only one part of the chain, is it not likely that the actual circuitry involved in audio playback (both digital and analogue) behave slightly differently, in terms of timing and jitter, with different encodings.

nb2

24-05-2010, 08:19

I propose a simple test to show how much data is "lost" by any lossy compression method, MP3 or AAC or Vorbis etc.

If we make the reasonable assuption that FLAC (or ALAC) does a fair job of compressing loselessly in the audio domain and that the resultant file has minimal entropy.

Take a raw WAV or FLAC. Convert to the desired lossy file. Convert back to FLAC and compare the size difference between before and after. That should give you a fair indication of how much actual non-redundant data has been lost as a result of the lossy process. I have not done this, yet.

I made a very quick test with only 4 files, and only three kind of mp3 compression
Even with this very incomplete and limited test, you get some clear conclusions.
Anyway it would be very interesting to make a more extensive test with a great number of sonically different files, and more MP3 kind of compression.

The test is not perfect, reconverting to flac can give a slightly bigger flac than the original flac.
Nothing surprising nor abnormal too.
Anyway, in my opinion, your reasoning is much more sensible, has much better logical support, and results are certainly more instructive.
As I thought, we are very very far from 80% difference.
A sonically ruined file being slaughtered by mp3 64 cbr transfer, result in less than 15% difference in flac file size, estimating data loss, for my 4 files.
With 80% data loss you probably could hardly recognize the song.
Very good mp3 vbr leads to very small difference.
All that is consistent with MP3 and FLAC compression principles, psychoacoustic, and listening ...

One point which doesn't matter, but we talked about that :
This actual test with only 4 files, shows that MP3 file size difference is NO WAY consistent with original and resulting FLAC file size difference, and obviously even less with data loss.
Try more files if you like, this inconsistency can only get worse, certainly not improve ...
Sonic content has probably the biggest effect in those differences.
This is quite intuitive if you know how MP3 and FLAC compression works.
If it is not intuitive, just pick up different songs with different sonic content, take your favourite MP3-FLAC converter, and try the test.

And most important, check your personal perception of those files quality, foobar is your friend.

Themis

24-05-2010, 09:04

A sonically ruined file being slaughtered by mp3 64 cbr transfer, result in less than 15% difference in flac file size, estimating data loss, for my 4 files.
You visibly misunderstand how MP3 compression/decompression works.

When you convert MP3 to PCM your decoder creates inexistent samples to make up for the sample rate differences (to PCM). The 15% loss in file size is just the way the "noise" (inexistent samples inserted afterwards) were compressed. It simply means FLAC compressed noise better than the music.
Your test proves nothing as to how much information is effectively removed in the MP3 compression process. ;)

By the way, you seem to think that VBR has better quality for the same resulting file size.
You should perhaps ask yourself the question as to why CBR or ABR exist, if the above assumption was true ?
Do you think that the codec designers created a less effective compression algorithms on purpose, just to mislead the consumer ? :eyebrows:

Peter Galbavy

24-05-2010, 09:23

By the way, you seem to think that VBR has better quality for the same resulting file size.
You should perhaps ask yourself the question as to why CBR or ABR exist, if the above assumption was true ?
Do you think that the codec designers created a less effective compression algorithms on purpose, just to mislead the consumer ? :eyebrows:

There are differing applications where tradeoffs must exist, for example encoding efficiency and real time latency - think streaming of live performances. VBR encoding is in strict quality vs. bit rate terms generally the best but it requires more data blocks to work which in turn increases lag. CBR is a much better bet for this application.

For why different types of encoding / queuing exist I suggest looking at VOIP and ATM (Async Transmission Mode) systems. VOIP is a classic for tradeoffs between psychoacoustics and transmission systems.

nb2

24-05-2010, 09:34

There are differing applications where tradeoffs must exist, for example encoding efficiency and real time latency - think streaming of live performances. VBR encoding is in strict quality vs. bit rate terms generally the best but it requires more data blocks to work which in turn increases lag. CBR is a much better bet for this application.

I'll add that cbr is often considered as obsolete.
The main reason why it still exists is because some early decoders you can find in old devices cannot read abr and vbr (or are buggy for those).

Nowadays, you should forget about cbr, and choose between abr and vbr according to your needs.
(abr being the natural replacement for cbr)

This is clearly explained in wikipedia and lame documentation for those who want to check.

Themis

24-05-2010, 09:58

There are differing applications where tradeoffs must exist, for example encoding efficiency and real time latency - think streaming of live performances. VBR encoding is in strict quality vs. bit rate terms generally the best but it requires more data blocks to work which in turn increases lag. CBR is a much better bet for this application.
Peter the important part of my question was : "for the same resulting file size".

I'm not discussing VBR vs CBR vs ABR, as these methods produce different file sizes. So the methods are clear.

Now : for the same resulting file size, do you pretend VBR (or whatever else) is able to produce consistently (ie very often) better quality ? :eyebrows:

Themis

24-05-2010, 10:10

Nowadays, you should forget about cbr, and choose between abr and vbr according to your needs.
(abr being the natural replacement for cbr)
BS. It's not what the tests prove. Look here : http://soundexpert.org/home

Peter Galbavy

24-05-2010, 10:40

Now : for the same resulting file size, do you pretend VBR (or whatever else) is able to produce consistently (ie very often) better quality ? :eyebrows:

If you use language like "pretend" then I will not bother answering.

Themis

24-05-2010, 11:28

If you use language like "pretend" then I will not bother answering.
I apologize Peter. I meant "claim".
Sometimes French words make it through. :(

Marco

25-05-2010, 05:24

Indeed. Peter, please make allowances for the fact that English is not Dimitri's first language.

Marco.

nb2

25-05-2010, 17:36

dave2010

25-05-2010, 22:08

Is someone interested in quality vs file size for MP3 CBR, ABR and VBR, as far as his library is concerned ?
If yes, taking a few files, actually convert them in MP3 and check sizes, is a process which takes a few minutes.
I have done it, result is valid for my MP3 files.
Do it, result will be valid for your MP3 files.
It is more efficient that listening to smart guys endless theoretical advice, mine, Peter's, or anyone else ...
This is a reasonable idea, though I'd suggest that if anyone has enough space that compression should be avoided, or if it works FLAC or ALAC should be used. For compressed music, then high quality AAC or MP3 should give better results, with AAC generally being better than MP3. Having said that, at high bit rates there may not be so much difference. Other issues also come to the fore, such as compatibilty with playback equipment, compatibility with other users. In the end it's up to each person to decide.

I have listened to many different compressed versions of music files I have over the years. It's an interesting exercise, and one can learn the characteristic artefacts of some compression methods. However, if you only want to listen to decent music, then it's not strictly necessary, particularly if you have enough storage space to avoid the need for severe compression.

Regarding how much data can be discarded, I'd say it depends. Going back to image analogies, if a high quality photo is taken, and then every other pixel is deleted in both the horizontal and vertical directions, you'll have discarded 75% of the data. You can also smooth out the image, by blurring, to fill in the missing pixels. The chances are it will still look pretty good. You might be able to repeat this several times, and still be able to recognise the contents of the image. Eventually there would be clear quality problems, but it would take some time before the subjects depicted in the images would become unrecognisable. This should at least demonstrate the possibility that discarding a lot of data in a digital representation can nevertheless still leave sufficient data to be informative.

MP3 or AAC compression really does discard a lot of data - probably around 80% compared with an uncompressed original sound source. Even more is discarded if the lossy compression is carried out to greater degrees of severity, such as Musicam/MP2 compression at 128kbps or 192kbps as used by DAB radio stations, or even further if aac+ (now AAC-HE) is used - as in some internet radio stations, which give arguably acceptable results even at 32 kbps.

nb2

26-05-2010, 14:53

MP3 or AAC compression really does discard a lot of data - probably around 80% compared with an uncompressed original sound source.

If 320 Kbps are available, from CD Audio data, you can keep losslessly everything for frequencies below 5Khz, and discard everything above.
(But this is such a stupid algorithm that I doubt it is used).
Do you really think that frequencies below 5 Khz are less than 20% of your audio data ?

(Anyway we all agree that for us, there is no reason to use lossy files when lossless ones are available)

Themis

27-05-2010, 05:00

Anyway we all agree that for us, there is no reason to use lossy files when lossless ones are available
Exactly, well said ! :)

In my case, I use lossy codecs for sonically degraded environments (crappy car
stereo, cheap ear buds). Synchronizing smaller files to the iPods saves a lot of time. ;)

If 320 Kbps are available, from CD Audio data, you can keep losslessly everything for frequencies below 5Khz, and discard everything above.
(But this is such a stupid algorithm that I doubt it is used).

If I recall properly, frequencies above 16kHz are almost entirely stripped away.
Can't remember in which codec, though.

dave2010

31-05-2010, 15:20

nicolas
If 320 Kbps are available, from CD Audio data, you can keep losslessly everything for frequencies below 5Khz, and discard everything above.
(But this is such a stupid algorithm that I doubt it is used).
Is this "fact" based on practical experience, or is there an information theoretic law lurking in the background?

Do you really think that frequencies below 5 Khz are less than 20% of your audio data ?Not necessarily - not really sure. In lossy compression the name of the game is moving the noise around to perceptual regions where the noise has relatively little effect. I still don't understand why you don't accept the simple arguments based on data size, which have real benefits both for storage and communications. There have also been any number of studies which suggest that most people can't tell the difference between lossy encoded music and the original using bit rates at 320kbps.

(Anyway we all agree that for us, there is no reason to use lossy files when lossless ones are available)I'm pretty much in agreement with this, as long as storage space is available, and the lossless decoding is smooth enough. Most lossless encoding does lossy encoding first, then uses algorithms to encode the residual error from the original. It perhaps then follows that the encoding and decoding are slower than lossy encoding/decoding.

BlueMax

06-06-2010, 07:04

Has anyone compared Spotify with MP3 rips on your HD?
I did some A-B comparisons of the same music.
I don't know which bit rate or format Spotify use but found that Spotify is better!
Strange! Any thoughts ?!

CDex settings that I use for MP3 rips are:
Lame MP3 Encoder
Bit Rate: 192 kbps min - 320 kbps max
Quality: High (q=2)
VBR Method: VBR-ABR
VBR Quality: VBR 0

Maximum

06-06-2010, 12:03

Think it's OGG Vorbis they use on spotify. Many people reckon it's better to mp3 at a similar bitrate.

Just noticed they have an open version now with 20 hours a month free and ads on the player.

Beechwoods

06-06-2010, 13:20

The Ogg Vorbis bitrate for Spotify free stuff is 160kbps. The premium stuff is at 320kbps.

BlueMax

06-06-2010, 13:51

Thanks for the info..
I have ripped MP3s for my MP3 Player but never tried OGG Vorbis.
Spotify and OGG Vorbis sounds fine; far better than digital radio.

Now that 99% of the people have broadband, 8MB and more, isn't it time someone offered live streaming of lossless files?

YouTube works fine in HD; so why not hi-fi audio?
Sure, with the potential of having infinite amount of music to choose from, it will be the final blow to the CDs. But recording companies should be able to raise revenue from streaming instead.

nb2

06-06-2010, 22:34

Is this "fact" based on practical experience, or is there an information theoretic law lurking in the background?

CD audio format is straightforward streaming 2 channel, 16 bits samples, at 44.1 KHz (needed to describe frequencies up to 20KHz)
that is 2 x 16 x 44.1 = 1411.2 kbps (CD audio bitrate)

With exactly the same scheme (which would be absolutely stupid, there is not even the slightest lossless compression), 2 channel, 16 bits samples, but only 10KHz sampling (which describe frequencies up to 5 KHz, well, actually slightly less, but I rounded)
2 x 16 x 10 = 320 Kbps

Ok, let's keep it simple :
With 1411.2 kpbs, you describe 2 channel, 16 bits samples of frequencies up to 20 KHz.
With 320 Kbps, keeping 2 channel and 16 bits samples, you can keep about 25% (ok I rounded 320 / 1411.2) of those frequencies.
Forget everything above 5KHz, and you are done.

This is a perfectly stupid and inefficient way to generate audo files, so I have no doubt that any compressed file format at 320 Kbps gives a much better result.
Keep in mind that 20 KHz is overkill.
Even a teenager with bat ears cannot hear it, or howls with all cats and dogs around if he does.

Not necessarily - not really sure. In lossy compression the name of the game is moving the noise around to perceptual regions where the noise has relatively little effect. I still don't understand why you don't accept the simple arguments based on data size, which have real benefits both for storage and communications. There have also been any number of studies which suggest that most people can't tell the difference between lossy encoded music and the original using bit rates at 320kbps.

Because my actual knowledge of compression principles always showed me this is wrong.
Use different lossless algorithms to compress data, you will find very different compression ratio for the same data.

Look at the spectrum of some of your recorded music, especially if you discard what is below -30 db , you will understand why it is so hard to pick up mp3 320 from lossless flac, and why 80% of discarded data is not realistic.
You can also find interesting info about fundamental ranges of music instruments there.
http://en.wikipedia.org/wiki/Range_(music)
Very few of them can reach C8

As far as I am concerned, what I still don't understand, is why everybody is so interested in theory, and why apparently nobody is very interested in sharing actual experiment with real music tunes ... :)

dave2010

07-06-2010, 07:00

Use different lossless algorithms to compress data, you will find very different compression ratio for the same data.This is certainly true for MP2, MP3 and AAC, with - IMO - the best sounding results using AAC for a given resulting file size.

The encoding quality parameters should not be ignored either.

Look at the spectrum of some of your recorded music, especially if you discard what is below -30 db , you will understand why it is so hard to pick up mp3 320 from lossless flac, and why 80% of discarded data is not realistic.Are you saying here that if you discard signal input below -30dB, and then do the encoding to MP3 320 and to FLAC that it would then be difficult to tell the difference? For classical music this might be silly anyway, as some music is very quiet. If you are saying to do this with music which contains several instruments (for example) and then filter out those parts of the signal which represent quiet sounds, or instruments playing quietly, that there won't be a perceptible difference in recording to FLAC or to MP3, then I'd agree with you. That's the whole point of lossy compression - sounds which are quiet and masked by louder sounds are simply removed, as most people can't hear them anyway.

Note also that it's possible to get reasonably acceptable compression on some music files simply because the spectral content is really not that challenging. I've heard some excellent guitar music from Napster done at modest mp3 bit rates. The challenges occur with much more complex musical passages in which either there is simply a clear loss of data/information perceptible to those who know, or, perhaps worse for most people, audible and unpleasant artefacts from the compression process.

For examples of aac+ (AAC-HE) try http://www.tuner2.com This should demonstrate that low bit rate encoding can be acceptable at one level, but also that it is not really high quality.

You can also find interesting info about fundamental ranges of music instruments there.
http://en.wikipedia.org/wiki/Range_(music)
Very few of them can reach C8 Thanks for reminding us of this. One obvious corollary to this is that this shows that the frequency range above 4kHz mostly gives us timbre information, allowing us to distinguish between a clarinet and a violin or a flute etc. Some instruments, such as the cymbals and gong don't just produce pitched sounds though, and I think that cymbals and gongs are very wide spectrum sources. A jazz cymbal played with a brush probably has quite a lot of high frequency content, though not necessarily more than a simple "clash" of two cymbals brought together.

.... is why everybody is so interested in theory, and why apparently nobody is very interested in sharing actual experiment with real music tunes ... :)You may be wrong there. I've been comparing different compression systems for years, and I think others have too. Keep coming back at us though. It helps to clarify issues.

Some are now noting that we can do most 2 channel audio using lossless formats or lossless compression, as we have large storage devices or high capacity media or use high capacity networks. Unfortunately, there may still be an interest in lower quality audio for broadcast purposes for some while.
Encoding schemes such as aac+ (AAC-HE) give passable results with very limited data streams, but they are significantly inferior to encoding methods at higher bit rates. Broadcasters often do not care about this, as they think that most people listen in their cars, or on cheap equipment anyway, and for commercial reasons they often prefer to have more channels rather than fewer high quality channels. In one sense there's not much wrong with this, but the danger is that future broadcasting (e.g DAB or DAB+) will completely rule out high quality channels. Personally I'd rather have a load of lowish quality channels (for car listening, everyday use etc.) and a few really high quality ones, but I doubt whether a lot of broadcasters will see things that way. It is going to be worthwhile hanging on to CDs, SACDs and in some cases DVDs (DVD-As) for some time to come.

nb2

07-06-2010, 07:41

Are you saying here that if you discard signal input below -30dB, and then do the encoding to MP3 320 and to FLAC that it would then be difficult to tell the difference?

No, I am saying that I don't believe that 80% of the data are in very low level ( and in harmonics above 5 KHz ) ...
Harmonics is important information, but I think you have much more than 80% of data in main information below 5KHz (which is very very high frequency, as far as human ear is concerned)

Check the spectrum of a few seconds of a normal level sounding music, with Audacity software for example.
You usually have material at -30db, even -60 db in those few seconds, mixed with 0db data.
I am quite sure it is very difficult to notice their removal.

That makes me understand why mastering is so much more important than slightly lossy compression.
As a matter of fact, mastering IS lossy compression (not only that, but it is) , and far more destructive than mp3

I am speaking for 16 bits samples used in CD.
24 bits samples are certainly such a huge step higher in quality, that I guess that anybody interested in high quality should consider the CD is dead.

nb2

07-06-2010, 07:52

I've heard some excellent guitar music from Napster done at modest mp3 bit rates. .

Yes, I was told that guitar easily sounds very good, and should not be used to check the quality of a system.
It is often excellent in movie theatres.

dave2010

07-06-2010, 21:48

Nicolas

One can only imagine what goes wrong sometimes. Currently listening to Puccini - Preludio Sinfonico and Capriccio Sinfonico played by Claudio Scimone and the National Orchestra of the Opera of Monte Carlo. Disgusting sound via Napster.

Yet it can't just be the music complexity, as a few minutes ago I listened to
Richard Chailly playing exactly the same pieces also via Napster with the Berlin Radio SO. Somewhat better sound, though still rather raucous in places. Presumably somebody wasn't doing a quality check for the Monte Carlo version, which is rough throughout, when they made the MP3.

The Berlin version isn't by any means good, but is relatively less disgusting.

MartinT

08-06-2010, 06:59

nb2

08-06-2010, 17:03

One can only imagine what goes wrong sometimes.
Right, it is very hard to understand where comes the difference from.
Is it the same mastering ?
How radios or Napster performed the encoding ?
I am convinced good 320 kbps streaming can be almost impossible to pick up from CD, but you don't have often such bitrate, and with low bitrate, that is another story ...

Marco

09-06-2010, 09:52

The dilemma is that CD may have a technically archaic encoding format, yet the best CDs played on a high end player can sound absolutely magnificent. There is still more to be learned about the effects of digital artifacts.

Hear, hear! :clap:

Your last sentence is almost certainly a truism.

Thing is, Martin, how many people do you think own a genuinely top-notch CDP these days and are therefore au fait with what CD is really capable of?

Most people these days use mass-produced players with plastic DVD-ROM mechanisms (or even worse, DVD players!), cheap PSUs, and sub-standard build quality, both externally and internally, so how they can hope to know what CD is genuinely capable of, sonically, by using such inferior machines is anyone's guess! :rolleyes:

The fact is most people today haven't heard CD done properly, and thus have embraced computer audio either through desiring the convenience factor, wanting to be up-to-date with new technology, and/or because they've heard some of the bland, tinny sounding monstrosities currently on the market posing as CD players, and then compared those to music streaming via a computer and decent DAC, and found the latter infinitely more preferable, simply because the quality of their CDP benchmark was so poor.

Now don't get me wrong, computer audio done well is phenomenally good and can more often than not sonically outperform even the best CDPs one can buy, *but* in my opinion there exists a misconception these days that any half-decent music streaming system will automatically outperform ANY CD player, which when you've had the relevant experience of using a genuinely hi-end machine, and compared the difference between the two on many occasions, is quite patently bollocks.....!

<Rant over>

Marco.

dave2010

09-06-2010, 10:44

Nicolas
Right, it is very hard to understand where comes the difference from.
Is it the same mastering ?
How radios or Napster performed the encoding ?
I am convinced good 320 kbps streaming can be almost impossible to pick up from CD, but you don't have often such bitrate, and with low bitrate, that is another story ...After years of listening I still don't fully understand this. Some of my earliest streaming audio listening was to Radio Sweden P2, at 96 kbps and probably using Real Audio codecs at the time. I was aware that it wasn't ultra high quality, but it wasn't bad either. Later I found stations such as Belgium Musiq3 - which was outputting at 128kbps mp3, and that sounded quite good. The BBC DAB broadcasts use Musicam (MP2) for their encoding, and Radio 3 is usually at 192 kbps. A few years ago they tried to reduce the rates to 160kbps, and many people noticed the quality loss. They excused themselves by saying they had new codecs, and that the quality should improve. They reverted back to 192kbps. There are internal BBC reports which suggest that it's been known for a long time that the bit rate for DAB should be 256 kbps (or higher if possible) for good quality, but like most broadcasters they have reduced their rates. My view is that 192 kbps MP2 is more or less equivalent to something between 128kbps and 160 kbps MP3, and that 128kbps MP2 (Radio 2, Radio 4 etc.) is more or less equivalent to 96 kbps MP3. ClassicFM is the only other UK radio station which uses a relatively (!!!) high bit rate of 160kbps using Joint Stereo, but they manage horrible sound quality anyway.

There does seem to be something in the quality of encoding - some encoders (perhaps with more carefully chosen parameters) do do better at low bit rates. Most do a lot better at high bit rates, and until recently I'd assumed that anything above 256 kbps should be good with any codec. However the recent post about Spotify and Ogg Vorbis has made me rethink this, and I need to investiigate that further.

There could be several reasons for poor quality compressed audio, which I hear sometimes from sources such as Napster.

1. The encoding is just not done carefully enough.
2. The original source has out of band frequency components which interfere with the encoding, and produce even more audible artefacts than those from a source without out of band components.
3. The "original" is in fact itself also in a compressed digitial format. The FAQ for Ogg Vorbis makes it very clear that because OV compresses in a different way from other methods, such as MP3, that compressing (say) an MP3 source to OV causes much greater loss, with consequent audible disasters. The BBC has also noted problems with transcoding music from one compression scheme to another.
4. The decoding resolution is 16 bits. [can happen sometimes ...] Interestingly MP3 can handle up to 24 bits and at different sampling rates. I think most players use floating point arithmetic. Playing back MP3s through a 16 bit D-A can, I suspect, give lower perceived quality - though this might only be a slight effect. I do have a hunch that it can take the edge off a "good" MP3 recording.
5. The bit rate is too low.

Although low bit rate compressed music can sound bad, and sometimes this is inevitable, I submit that this (low bit rate) is often not in fact the reason why it is terrible, as sometimes with care it is possible to do a decent compression at relatively low rates. Probably systems such as Napster don't have any really strong quality control procedures, so there is no attempt to optimise quality. If everything is put through the same encoder at a fixed low bit rate, then it is quite likely that the results will sometimes be horrible - as noted.

Note one thing, which I think has confused people. Most digital compression schemes these days cope with wide dynamic ranges for the source material. Some compression schemes, such as NICAM, address dynamic range explicitly by normalising, but this only gives a few dBs of improved signal to noise ratio. Discrete Cosine Transform methods, such as MP3 or other sub-band coding methods, handle wide dynamic ranges simply by what they do to the coefficients in each time window sample, so there should not really be a problem with loud and quiet sounds.