I recently discovered Musepack and have started ripping my CD collection to MPC so I can have a "virtual jukebox" and as backup so I can box up my collection. Anyway, I'm quite happy with the quality but now I'm wondering if I should just go with a lossless compression. I really like the looks of FLAC and think I'll go with that format if I go lossless. I'm just wondering if the extra space I'll need will make it worth it in the long run.

Ugh, decisions, decisions. Just looking for some opinions out there.

I was the same as you, converting my collection to MPC.
Now I am seriously considering changing to Wavpack (get the best of both worlds).

Try a search and look into it, it's lossless but it also has a lossy mode, the lossy mode creates a correction file which can be used to re-encode to the original wav any time you want (using the lossy file+ the correction). It's a brilliant concept IMHO , look into it as an alternative to full lossless.

Consider the size of your collection. I have well over 1000 albums. I could never afford the space needed for lossless, Musepack only for me...

Making backups should also be considered. You need a backup system with massive capacity to store all that data.

Even with backup it would be a pittance compared to what you paid for the albums....
Disk space to hold a Flac (or whatever) compressed album is probably about 25 cents US
with the cost of HD going under $1/GB.

Backup to DVD (just media cost) about 5 cents?

It sound neat, but maybe i'm just missing the real advantage. Unless you keep alll your correction files
in some off-line storage, you don't really save any space, do you? I guess if you use remote storage
you save network bandwidth when playing the lossy portion.

One rule of thumb is that a change isn't worth making unless it gains you something in the neighborhood of an order of magnitude; so I don't use lossless compression at all. A factor of two in storage space just isn't worth the effort and risk of encoding and decoding -- you might as well keep the wavs.

I encode using mpc --standard --xlevel, which does give close to an order of magnitude compression, which is transparent to my ears, and which encodes and decodes very rapidly.

Of course adherents of lossless compressions will disagree, and that's fine; I wouldn't expect everyone to agree with me, this is only a rule of thumb, not a law of nature.

>you might as well keep the wavs.

With waves though you cannot add ID tags.

That's what I was going to say. And on top of it the space cost is more than sole lossless space cost because of the lossy file. I don't see no reason how the difference file could take less space encoded in lossless rather than the original file. Thus if you don't move your files around I don't think WavPack has an advantage. Because space is so cheap I am using "--braindead --ms 15 --xlevel", lossless is an overkill for me at the moment. Maybe in ten years...

I love these quality arguments.

I have been reading in a few threads about lossless files becoming corrupted when backed up to cd. That would throw a wrench into the whole archiving concept.

I use aps. It seems to be transparant for everything I encode.

Artifacts are rare, and it seems that most are brief, and you need to carefully compare the original to the encode to pick up the differences in most cases.

I don't do this because the last thing I want is to become an expert at picking up tiny flaws in my collection.

I often wonder if the super high quality/high bitrate encodes mostly serve to give the user that "warm and fuzzy" feeling that they have done the best that is technically possible.

I use flac for the albums that are very important for me. The ones I like somewhat but won't listen to regulary are fine off with mpc -xtreme.
Both encoders are very fast btw.

With the .WAV specification, you can stick whatever the heck you want after a .WAV file and it will still decode properly, assuming the decoder is operating to spec. That's just a stupid player limitation. I wonder if foobar overcomes this.

I believe it is.

* It is possible that a Musepack SV7 player will be very hard to find in 10-20 years.

* Storage space gets cheaper and cheaper. 300 Gb will not be much in 5 years.

* You can always encode a lossy version for a portable device without transcoding.

* In some years, Room Correction might be a reality for normal people. Room Correction might work better on lossless files, but that is only a guess.

* You will never have to worry about the quality.

seezar:

If you've got the space (and I mean, like, LOTS of space), go for lossless. But keep in mind that even though storage shouldn't be too big a problem with large hard drives, creating backups can often still be a pain - and "comfortable" solutions, such as a second hard drive that you regularly transfer your whole collection to, are quite expensive. (Not to speak of DVD burners... )

If you haven't got the space and want to stay lossy, then make the most of it - i.e., don't waste your disk space using settings higher than --quality 6. The additional space used by lossless is justifyable just because of that, because it's lossless. The space wasted by insanely high MPC settings isn't.

Edit:



There will be a tool to convert SV7 to SV8 when SV8 is out. MPC playback is nothing to worry about as far as I'm concerned, with the decoder (and possibly the encoder, in future) being Open Source.

I used to used high-bitrate MP3 but now everyhing I do is lossless; also starting to use FLAC more than APE after encountering some corruption on freshly done files (most likely system issues - need a reinstall badly right now), but looking at Optifrog as well. Cross-platform is beginng to matter much more to as well.

Speak for yourself, Dude What if you have the space but use lossy? Why is the space using MPC "--braindead" rather than "--insane" any more wasted than the lossless? That's your personal conviction. I am happy with those wasted bits. It's all about how safe you want to be. And that all depends, so you be the judge.

Allot of Microsoft example code will fail to load a wave file if it is so much as 1 byte over the internal length - if you were doing it propperly you would write a different 'riff' chunk, but nothing would touch it...

atici:



Then that's your decision. But actually wasting space (which is, IMHO , what you're doing) "just because you've got enough of it" doesn't seem very economic to me. If my fridge is loaded full with food, I don't eat all of it "just because it's there".




I was talking about the file size increase of --quality 7/8 over --quality 5/6 (which is >30% in most cases, going from --quality 5 to 7). I think the additional space is wasted because you don't gain any practical advantage over using --quality 5 or 6. As some people have already pointed out, there will always be cases where a lossy codec (whatever the bitrate) will not be transparent. You're never 100% safe with lossy codecs, that's a risk you've got to take - pumping up the bitrate to crazy levels doesn't change that at all.

With lossless, you are totally on the safe side - you've got a 1:1 copy of the CD. A lossless music library is future-proof, flexible (if you decide to switch to superior format, you can just DO it), convenient (for transcoding to lossy for portable use, or for friends who prefer different formats, etc.), and best of all, you don't have to worry that anything could be wrong quality-wise. All of that justifies a large increase in space (if you can afford it), IMHO. If I had a large hard drive, I'd definitely go for it (it would save me thousands of runs to the library every time I decide I want a 1:1 copy of a CD I didn't archive losslessly before ). Unfortunately I haven't, so I've got to stick with MPC --quality 5.

What i would do in your situation:

Got the space, go lossless. That way you never have to worry if you have chosen the right codec, setting or anything else. Lossless is lossless. If you run out of space you can always transcode to musepack.

What i'm doing in my situation:

Rip the cd's i own myself to musepack standard. I keep my cd's nice and clean so i don't really worry about unreadable/damaged discs in the future. I had a big hd crash sometime ago so i don't really rely on hd backup anymore.

Rip the cd's i do not own myself to lossless (monkey boy for me) and store that on cdr. That way i can always do myself or somebody else a favor and transcode to another format if that's needed.

I got tired of worrying about finding the perfect lossy encoder, settings, and build version, and started over with FLAC. I have a bit over 750 rock/pop CDs on one Maxtor 250 GB drive in my music server, and a second drive waiting to be filled with my Celtic, Blues, Jazz, and Claasical CDs.

With FLAC I have source code and no licensing worries, so when I get around to it I can write custom software for things like automated transcoding to MP3 for my flash player, and streaming to a wifi-enabled PocketPC and Home Theater PC.

If you can afford it, I'd say go with FLAC lossless.

Just backup the correction file to cd-r, then you'll save space as you will only have the lossy file on your hard disk but you will always have the means to re-encode to the original, unless your cd-r's go pear shaped , but then there is the chance of file loss with any backup medium.

Look for some of Den's posts, he has done a lot of testing with Wavpack and says that the lossy file is more than good enough for transcoding to other formats, eg MP3, Ogg Vorbis etc.

http://www.hydrogenaudio.org/forums/index....2&t=8416&st=25&

If lossless could compress files small enough to fit 2 average CD's on a cd-r (for backup) then I would use FLAC but as it is just now you are as well making a copy of the whole CD uncompressed, from an archival perspective.

As usual everybody will have their own pro's and cons for using/ not using a specific compression format and the debate will probably never end. Like I have said before it's all down to the individual, but it's great to have a choice. B)

Why can't we apply the same idea with any kind of lossy format? Isn't there a general difference calculator that we can pass through any lossless format of our choice that could be combined in the player plug-in? I think it's technically feasible. We also need a command line joiner to complete the toolset...

Interesting point, MPC lossy with a correction file to re-encode the original.

I did read on another thread (but can't find it now) someone suggested the same kind of idea with a combination of Ogg Vorbis and FLAC.

About correction files:

If you can take the original wav, encode it, decode it again, then paste the inverse of the decoded wav into the original wav, you'd have the encoded file and a correction wav.

All you'd have to do is paste the correction wav back into the decoded wav to get the original wav...

Someone also mentioned this on another thread about WavPack lossy, and included something about a slight increase in efficiency (1-3% I think, something minor) by using WavPack Hybrid + Correction over MPC + correction.

Well I'm sure it must be very easy. Two command line tools would solve it, one to calculate the difference another to sum them. But if we want lossless playback in the player side we need plug-in support. I just know how lossless encoder works vaguely, but I think there's still probably redundant information in the scheme and the size of the lossless difference could be even lowered if one has information about the lossy algorithm. It will be splitting one lossless encode of the original song into two basically. Thus the space requirement would be only as much as the regular lossless encode (ideal case).

How would a second order lossy pass work? Take one wav, do a mpc lossy encode, look at the difference, do another lossy encode and do a lossless encode of the second order difference. With these two command line tools our options are infinite. And maybe we can get a better compromise in quality/space cost.

BTW with Wavpack, why do you lossless encode the difference instead of the original. Because the difference is useless without the lossy part and lossless encode of the difference vs. lossless encode of the original do not have a difference in filesize (or do they). So you can as well keep them as MPC files and FLAC files in separate folders and your space consumption will be the same with playable lossless files. But if you can lower the amount of redundant information and make the lossless encode of the difference file smaller, then that would be something!

I am not very up on the tech side of all this, but doesn't Wavpack do all of this in one go, encode to lossy+correction and then re-encode to wav using same. If MPC (or any lossy format) was modified to save the difference as a seperate file during encoding (as an option) then the de-coder could just use the 2 files instead of one to re-encode to the original wav.
In essence, isn't a lossy encoder/ decoder already performing the basic tasks required to do this, it's a case of saving the difference during encoding and then re-applying it again during de-coding.
I'm probably making it sound over simple in my ignorance, but like I said I am new to this.

Yes, wavpack does this all in one go. In the other thread, someone made a MPC, then subtracted the difference between this and the original and made the difference into a FLAC.

In theory it should work, but IIRC there was a number of checksum errors when they tried to recombine them...

The other aspect of Wavpack is that the lossy file is very, very good. While some don't like the fact that is has no psychoacoustic modelling to improve the use of bits, I prefer this because you don't get artifacts like the others. It also has zero pre-echo, and encodes the full spectrum. The latter might sound wasteful, but I'm noticing much better treble, particular with high end percussion in my music, largely due to the lack of pre-echo or smearing of transients that you get from the other lossy codecs.

Also, my tests showed that Wavpack lossy is superior to transcode from compared to other lossy formats. I compared transcoding from Wavpack lossy 320, LAME insane, MPC beyond braindead and Vorbis q10. Wavpack was consistently the best for me, with zero introduced artifacts from transcoding. All the others added something when transcoded into ATRAC3, MP3 128kbit and Vorbis q4.

Wavpack, is also freaking fast to encode/decode so it is very convenient to use for both lossy and lossless encodes/decodes and transcodes.

I'm now combining it with PAR2 for the correction files to get the best of both worlds, and reduce the chance of losing my correction files on cdr from cd cancer... I have high quality lossy on hand all the time, and lossless when required from cdr.

Den.

@Den

Yeah, I have to agree. Thanks for the work you did in regards to Wavpack, it was your threads that first switched me onto it. Really interesting

I had been looking for something better for transcoding to my portable payer and this was perfect without going full lossless. Planning to do some tests myself when I get home again.

Edit: because I posted it unfinished by accident.

This I must criticize now. Doing a transcode from an already lossy file (and Wavpack is a hybrid, I know), if you still have the absolutely lossless original, is always a bad thing, which should be avoided.
And boy, the future backup-medium is DVD. For more than 9 months now I´m backuping every Original-CD I get from friends or libraries to DVD. I had no dropouts since today with it. On average there fit around 11 to 13 complete lossless single-interprete albums on a DVD-R (compressed with Monkey´s Audio in "high"-compression mode).

@R.A.F.
Your criticism is absolutely correct, but I have my reasons. I now have 100's of CDs archived on my PC as Wavpack lossy. Every day, I go to my PC, fire up foobar, pick 2 hrs 20 minutes of music to listen to, then dump these out as wavs from foobar and encode/transfer them into my Minidisc for a day's listening. If I had all of these files as wavpack/flac/monkeys lossless, I simply would not have the space.

The above process takes ~10 minutes to do, outside of the time required for me to actually make up my mind and pick my favourites.

By using wavpack lossy I do have the space to keep everything handy on HD, and the thing is that wavpack lossy does not introduce significant artifacts when I transcode from it. Every other lossy codec I have tried does, even at ~320 kbits. The only one I haven't tried is AAC, but that's coming soon.

When I am compiliing something more important that just my daily tunes, such as burning a CD, I can dig up the Wavpack correction files from CDR, and get the truly lossless version. As you can imagine, to dig up all the correction files for different artists, transfer them back to the PC, dump them as wavs, etc, this takes a lot longer than 10 minutes.

The crazy thing is, that I find it very hard to pick any significant difference between the lossy and lossless versions anyway, and if I lost my correction files, I'm not sure if I would be all that cut up about it, because wavpack lossy is truly amazing to transcode from. Test it for yourself and compare against the other lossy codecs and see for yourself.

I also don't have access to a DVD burner.

Den.

Using standard lossless encoders (tried on Monkeys Audio) on files with a lossy component subtracted results in files which are larger (the correction part on its own) than if you losslessly encoded the track on its own, this is because the lossless predictors cannot predict for lossy subtracted data.

That is why I like wavepack so much, it goes that extra hurdle.

I made a lot of tests when looking for the best trade-off between lossy and lossless, as lossless is for me waste of space and lossy is a disadvantage, because of the fact you're losing something. The fact is, I bet the most people wouldn't discern any difference beyond 352 kbps WavPack lossy (high quality) mode... As you may know, I am a little snobbish (when the sound matters...) so I wouldn't use anything, that wouldn't meet my demands. So after making very hard tests I set the border for me between 448-512 kbps WavPack lossy (HQ) mode. Just to have some overkill and to be sure. This is enough for archiving and possible transcoding later. And transcoding is no death in this case, you can believe me You must have a real bat ears to hear any difference worth mentioning. As for classical music, I use WavPack lossless, as compression ration is about 2,5:1 in average, which is close to 512 kbps... So in the long run I don't need as much diskspace I would have needed with lossless or wavs and the quality is really "high" enough.
Just my two cents...