View Full Version : MP3 encoding options
Not sure if this should be in C&C or here, but meh. :)
I'm currently re-encoding my CD collection (for the umpteenth time) using ABR-256kbit encoding. At home I drive the audio through decent hardware - Denon and KEF in the office, Denon and Mission in the living room. On the move, I use CX 500 headphones. I'm happy with the quality.
However, this format seems to eat MP3 player batteries for breakfast. Newer players seem better than old ones, but still. So, my options are:
Carry on and live with it.
Re-encode (again) at CBR-320. 30% disk space hit, which isn't too bad. But would I notice the quality change, or any difference in battery life?
Use MediaMonkey to downsample to CBR-160 when copying files to MP3 players. I get high quality at home and decent quality on the move (when quality isn't so important anyway) with better battery life. If I do this, is the choice of source encoding (ABR-256 or CBR-320) likely to affect the quality of the downsampled files in any noticeable way?
Encode as FLAC and downsample. I'm no audiophile and I don't think I'll be able to tell the difference. I will however be able to spot the 300% disk space I've lost.
I don't know enough to answer this. Can anyone help?
PS - I do volume levelling and all that jazz. Audiophiles will cry foul but it saves me reaching for the volume control at every turn when I've got the player on shuffle.
8/10 people cannot tell the difference between 128k CBR MP3 and CD and 99% of people cannot score better than chance between 192k CBR and CD.
And thats in both a proper listening room and/or at home with their very expensive hifi setups :)
I'll let that snippet of info allow you to decrypt my feelings on the matter :)
Always used LAME with --alt-preset extreme here, sounds almost exactly like the CD, CD's sound slightly more alive when I use my Headphones and there is a lot going on in the music, but in all honesty it's barely noticeable.
I must be in that 1% - Anything under 256k just doesn't sound right to me, though a lot of it is dependant on what kit your listening to it on, on an iPod for example, I doubt I could tell.
I'm going to bet you're wrong and that its all in your head :)
Your brain simply cannot hear the missing frequencies even if they were there.
Pretty much what I thought.
I wouldn't be concerned except for the battery life issue. It is noticeable - the processor has to do more 'work' to decode VBR/ABR than it does CBR, and obviously memory usage goes up proportional to bitrate. Also, some iPods have problems with VBR/ABR though I don't know if that also applies to the iPhone/iPod Touch.
I really don't want to encode everything twice but I don't want crap results as a result of transcoding (in this case between 256kbps ABR and 160kbps CBR) either.
Thats a point I didn't make and perhaps should have. Those MP3s I referred to are all done properly - multi-pass fraunhoffer encoded MP3.
Transcoding is easily audible because the compression algorithms calculate the tone masking and temporal masking incorrectly because masking has already been applied once. This leads to poor file sizes and incredibly audible "errors", which get worse the higher up the frequency spectrum you go.
Avoid transcoding at all costs!
So how does one go about converting one's audio files from [say] Apple lossless to multi-pass fraunhoffer encoded MP3 files?
Lossless to MP3 is fine because the first pass is not lossy and no masking has been done. :)
Guess I'll have to rip twice then, or take the FLAC hit, or the battery hit. Ugh either way. :(
PS - I'm using LAME on the highest quality settings it does - CPU time is cheaper than my time. If there's a better encoder out there I'd be willing, especially if I'm going to have to re-do the 80 CDs I've already done so far. Rather do that now than have to do it all again with the 300-400 CDs I've got total.
I'm going to bet you're wrong and that its all in your head :)
Your brain simply cannot hear the missing frequencies even if they were there.
I wish there was some test you could do because I am pretty certain I could tell you the difference.
Flibster
03-08-2009, 22:33
I only get a couple of hours less playtime using FLAC than mp3. Still get over 40 hours with the D2 anyway unless I'm playing audible files.
I always use flac now though. I've got the disk space on the server free and as they get shuffed around the flat and played though various devices it's no hassle as they all play flac.
Downsample to 256kbps mp3 for the other half's mp3 player though Can't play flac unless I rockbox it - and she can barely understand the menus at the moment. ;)
8/10 people cannot tell the difference between 128k CBR MP3 and CD and 99% of people cannot score better than chance between 192k CBR and CD.
And thats in both a proper listening room and/or at home with their very expensive hifi setups :)
I'll let that snippet of info allow you to decrypt my feelings on the matter :)
8/10 people are deaf then. :p
Switching to FLAC isn't a huge issue for the home system as I've also got a server which I'm about to add 2TB additional storage too, but I do like to copy the library around (e.g. on to a laptop to take to work). FLAC isn't an option for that so I'd have to use the downsampling trick there too, which is a pain.
Whatever I do, it's going to be a compromise. Only one of my portable players supports FLAC so going all-FLAC isn't even close to being an option.
I wish there was some test you could do because I am pretty certain I could tell you the difference.
An ABX test would be the first place - foobar can do it IIRC.
Point it to a pair of 'identical' files but in a different file type and then it will play snippets of each and ask you to identify them.
It will then tell you if your results fall with the realms of guesswork/chance or actual ability to tell.
Description is a bit vague but thats the jist of it, been a while since I used it.
8/10 people cannot tell the difference between 128k CBR MP3 and CD and 99% of people cannot score better than chance between 192k CBR and CD.
That can't be true surely? 128K sounds awful.
Well, I've taken the hit and switched to FLAC. Got a MediaMonkey Gold license too so I can rip once and for all and transcode in whatever direction I may choose.
I'm still using CDex to rip the library because I've invested a lot of time in getting a 'just right' CDDB database and I don't want to do that again. I have changed my procedures though so the ripping process should be 50% faster and I'm applying ReplayGain to the finished files instead of volume levelling prior to encoding (no point using lossless encoding on an inherently lossy source).
That can't be true surely? 128K sounds awful.
I'm going to go out on a limb and say that the 128k you refer to was a poorly encoded MP3 that you got from the net or ripped yourself quickly on your PC using whatever software, rather than a carefully ripped 128k MP3 ripped carefully and properly with expensive software?
Sadly I have lost them now but I did have some test ISOs that you could burn, listen back and then tell me what was what for me to 'score' you. I did it on OcUK and I think only one person passed (rob.something IIRC) and the rest of the "audiophiles" seemed to fail miserably.
The same test was conducted in the multi-million-pound ISO standard listening room at uni, with top drawer kit in double blind conditions and again, people faired badly.
The science behind MP3 is extremely sound, and it does work. AAC is better at getting smaller file sizes for the same data, as is WMA, but the essence is all the same.
I'm with the Dr. on this. :)
Most MP3 encoding software seems to prefer speed over quality by default, and that's where the problems start. What's the rush? Ramp up the quality settings and let it take the few extra seconds. Of course, it does depend on other factors too - quality of the original rip, which encoder you use, and so on.
Anyway, my plan is now thus:
Update my homebrew tag validation software so it works without having to do multiple re-rips (done).
Rip to FLAC (highest compression setting).
Put a transcoder to work to build a portable device compatible library.
Consider what I do with the original FLAC files - either a higher bitrate transcode or leave them alone, depending on how big things get.
Due to the amount of crap I've got scattered randomly around all my PCs, I'm a bit disk space challenged currently. I really need to go on a crazy zapping spree and consequences be damned.
Sadly I have lost them now but I did have some test ISOs that you could burn, listen back and then tell me what was what for me to 'score' you. I did it on OcUK and I think only one person passed (rob.something IIRC) and the rest of the "audiophiles" seemed to fail miserably.
If only there was as a method as easy for testing people on their ability to tell cables apart - sadly you can't just download cables :(
vBulletin® v3.7.4, Copyright ©2000-2025, Jelsoft Enterprises Ltd.