Hi,

after consuming this excellent article (which made me change my mind in some points), I'm wondering if there's anyone who can add a "lossy comparison" with such a comparing table & stuff... after consuming some threads in this forum, I can't really decide between .ogg and .mpc anymore.

Take a look at this Vorbis [ogg]
AND THIS

EDIT: added another link

The disadvantage of the listening tests is that they don't comprise many formats at once... but they're a good source for a first glance, indeed.

You can always check the pros and cons for each format in their respective wiki pages: mp3, vorbis, mpc.

Judging by the page you originally linked you were actually looking for tables that listed features and that noted whether hardware/software support was good, what OSes there was support on, and so on ... that kind of thing.

I suppose this goes some way towards that:

http://en.wikipedia.org/wiki/Comparison_of_audio_codecs

It's the right direction, yes...
Some items like compression (well, on lossy codecs...) are missing there, though...

Come on, I'm kinda "newbie" ... I'm not that deep into technical things... yet.

I'm not that deep into technical things either, but I will stick my neck out here.

I *think* there hasn't been a major update of any of the comparisons between lossy codecs because, essentially, the war is over.
The contenders left are mp3 as encoded by LAME; AAC as encoded by Apple or Nero; Ogg Vorbis, especially the version tuned by Aoyumi.

The advantage of LAME is that everything plays it. The advantage of all the others is that they produce good results with smaller files than mp3. The advantage of Ogg Vorbis is that it is free and open source.

To get beyond that depends on either the specific hardware or software you want to use, or pretty fine discriminations on sound quality for file size, which only you can do, with your music and your ears; with your position on FOSS as a non-technical factor.

There, that's probably wrong, but it might provoke someone who really knows to put it right.

Best

Michael

But it's not conceivable which format will survive the next few decades?
I dislike transcoding into every new invention.

If you don't want to rip again, use lossless. The lossy format most likely to survice is mp3. If a lossy format is transparent to you now, it will still be transparent to you when you get older, so why bother switching?

I don't have a sinfully expensive HiFi set, so MP3 (well, I prefer .ogg right now, but its usage seems to be decreasing yet) is quite transparent, indeed.
I'm a perfectionist in some way. If there's sth better, I just can't ignore it. But maybe I should change my mind here...

MP3 has been the standard for more than a decade now, but times are changing... too many contenders out there.



I see... I'll need to wait for the next "big thing"...

I guess what most members here do is
- archiving losslessly so they are future-proof no matter where things go
- using mp3 for listening purposes other than on their pc.

You can exchange mp3 for ogg vorbis, aac, mpc, whatever you prefer resp. what's playable on your hardware. As you're a bit of a perfectionist you are supposed to use a bitrate of roughly 200 kbps on average, and with this you will get a comparable very high quality from any of these codecs. As you don't like reencoding mp3 is still the most future-safe choice IMO whereas with the other codecs you can save some kbps while still getting the same quality as with mp3.

An alternative can be to use a close to lossless archive, and use it also for listening purposes (and you have the option to use an additional correction file to fully restore the original version if you want that). You can do so by using lossyWAV together with FLAC (or use wavPack lossy, but I guess FLAC is the more future-safe choice which you care about). Close to lossless means using a bitrate of very roughly 400 kbps (depending on your demands). Transcoding is very good in case it should be necessary some day even when not using a correction file. Your playing machinery must be able to play FLAC however. IMO that's the pefectionist's best solution for listening purposes and a good solution for archiving at the same time.

One of the benefits of having a lossless archive is that my lossy library is inconsequential to me. It is the library that I use the most, by far, but it can easily be replaced, or updated to a new setting or format. It is well worth considering.

I wonder what the Next Big Thing might be?
CDs record two channels to the limit of human hearing. The best lossy encodes are indistinguishable from CD for, say, 90% of people 90% of the time, and for the remainder, Moore's Law has meant that lossless is quite possible.
That would seem to mean that the only improvement could be in providing more than the information in two channels. I am told that people with access to home theatre set-ups still find stereo good for music, so maybe surround sound on the present model isn't the NBT. Maybe something that more closely recreates the experience of being in a concert hall? But that would only be relevant for certain kinds of music.
And then again, two channels would seem to map pretty well onto two ears.
Of course, in the late nineteenth century, a lot of people thought that physics was more or less complete. But it looks to me that if there is going to be something really new, it will be a lot more radically transformative than a new lossy codec, and will involve new systems from end to end.

So ripping to lossless, and using whatever lossy is most appropriate that week should see you through for the next couple of decades.

Not that I know anything, but it all feels like mature technology.

I already try encoding with very high bitrates to avoid losing too many information (losing no information is impossible with lossy codecs, yeh..).

What is the main advantage of using lossyWAV in this case?




So do I. There's always sth which can be improved significantly.
New codecs raise and fall... TAK is coming, for example.




Currently I'm ripping to WAV (it has been the easiest way until now... you see, I never used FLAC or sth), encoding to OGG and deleting the WAV afterwards... so, at least, the second step is fine?
I'm gonna buy some new hard disk these days, however... so I should start keeping the original rips... in another format, of course.

(I hope I got it right now.)

The usual codecs like mp3, aac, vorbis internally do a pretty complicated transformation to describe the music. They use blocks of wave samples (called frames), separate the music into frequency subbands, analyze the musical content of the subbands with the target to omit what's not audible according to a psycho-acoustic model, transform the audible content approximately into the frequency domain (music can be described as wave samples which is a representation in the time domain, and it can be described as a superposition of frequencies which is a representation in the frequency domain), and store the frame's information of the frequency domain representation in compact form. There's more involved like taking into account the correlation between the left and right channel.

In the end as we all know we get excellent quality at a bitrate of roughly 200 kbps. On the downside the mechanism is so complicated that looking at the universe of music chance is there are musical spots where quality isn't top. That's what we call problem samples, and all the codecs mentioned do have problem samples though when using a good encoder they are rare and usually not annoying. That's why most people don't care about problem samples.
Using very high bitrate like 320 kbps with mp3 brings many problem samples to the inaudible or at least totally acceptable state. Hoewever you don't get perfection in any case even when going extremely high in bitrate. For instance I remember guruboolez having written here on HA that even with Lame freeformat at ~450 kbps IIRC he can still hear the inaccuracies of a castanet encoding.

In the end it comes to practically minded persons (most people) being totally content with mp3 etc. at a bitrate of ~200 kbps (or below). Moreover with a lossless archive in the back it's always possible to reencode a track in case it is found that the current encoding isn't perfect.

For the more or less perfectionist however the lossyWAV approach makes sense. The overwhelming difference is that the signal path is very simple without any subband coding and transformations. The background is that the reason why we have a 16 bit resolution with CDs is that the dynamic range from the loudest to the most quiet spots must be represented with a sufficient accuracy. At a quiet spot many of the leading bits are zero, that is the actual resolution is a lot less than 16 bit, and we still get a very good quality. With a loud spot we need the full 16 bit to differentiate from a quiet spot, but we usually don't need the 16 bit for a sufficiently accurate representation of the wave sample.
What lossy WAV does is analyze the music and round the wave sample so that a certain amount of least significant bits are 0. The number of bits set to 0 can be zero in case the full representation is needed for quality.
So lossyWAV reduces the accuracy of the wave samples representations. Within quiet spots it can do so in only a restricted sense in order to keep up an accurate representation, but with louds spots chance is high that a lot of bits can be saved.
The lossyWAV output is fed into a lossless encoder like FLAC which can take into account 0's in the least significant bits.
For the decision how many bits to set to 0 lossyWAV looks at the frequency spectrum and searches for low volume areas. The area with lowest volume decides on the number of bits to remove, in order to have these low volume frequencies still audible. This is the basic principle. A number of additional features have been added by Nick.C which make the procedure very defensive, and even allow for weakening the basic principle in an astonishing way arriving at pretty low bitrate (compared to what's possible with the pure approach) while still having excellent (though not perfect) quality.
Full transparency depends on the listener and is expected to be achieved in the 350-400 kbps range (on average) using quality options -5 to -3 according to personal demands. For the even more demanding listener quality option -2 or -1 can be used, especially if it's about archiving without an additional correction file or another fully lossless procedure.