[FAQ] Which format should I rip to?

[DreamTactix291]

Which format should I rip to?

Which format to rip to seems to be a question that is popping up a lot lately, so I'd say it's probably time to address this. There is no jack of all trades codec that is perfect for all reasons, but hopefully this will help narrow it down for some of you.

First I think it's necessary to explain what exactly it is we're doing when you rip and compress a CD. CD audio data is stored as uncompressed linear PCM (Pulse Code Modulation) with a sampling rate of 44.1kHz. Each sample is comprised of 16bits and 2 channels are samples. To be more direct, CD bandwidth is ~1.4Mbit/sec or around 10MB/min of music. This adds up really fast.

Next we need to narrow it down to whether you should use a lossless or lossy codec.

Lossless

• Ensures a perfect copy of your CDs at a compressed size
• Not nearly as small as lossy for most genres - average bitrate is around 800-1100kbps (950 or so for my collection) - roughly 2/3 of the original size of the CD.
• Can be easily transcoded to other lossless or to lossy formats without compounded transcoding loss also while keeping tags
• No lossless format is as prolific as MP3 in support, however transcoding to MP3 from lossless is extremely easy
• Recommended for people with large amounts of storage (>200GB)
• Increased disk usage compared to lossy on DAPs resulting in lower battery life

Lossy

• Never is a perfect copy of the source, although it can sound perceptually identical
• Considerably smaller than the source or losslessly compressed material. Perceptual transparency generally happens for the best listeners at around 190-200kbps VBR, while it may be as low as 130kbps VBR for less trained ones.
• Cannot be transcoded without compounding transcoding loss. Best to leave it in the lossy format of choice for good.
• MP3 is extremely prolific so transcoding with that format shouldn't be an issue.
• Useful for people with smaller amounts of storage (<200GB) or for portable use (such as a DAP)
• Reduced disk usage compared to lossless on DAPS resutling in better battery life

Now that we've explained the very basic differences between lossless and lossy encoding, now we break it down to different codecs.

Lossless codecs

FLAC - Free Lossless Audio Codec

• Asymetrical codec (decodes faster than encodes)
• Extremely well known and prolific codec. Highest amount of hardware support
• 9 different compression levels (-0 through -8), however there isn't much difference in all of them and decode speed is about the same for all of them
• Compression is fast but moderate in level, decompression is very fast
• Uses FLAC tags (which are identical to Vorbis comments)
• Error resiliant
• Has excellent cross platform support
• Currently still developed
• Open source

WavPack

• Symetrical codec (decodes at same speed of encode) with an asymetrical option using -x
• Not very well known codec. Has hardware support on some Rockbox enabled devices (iRiver H1xx and H3xx)
• 3 different compression modes, each with a different encode and decode time (fast (-f), normal (no switch) and high (-h))
• Comrpession ranges from very fast to fast and moderate to good in level, decompression ranges from very fast to fast in speed
• Uses APEv2 tags
• Error resiliant
• Has best support on Windows, but is slowly gaining support on other platforms especially Linux
• Currently still developed (very actively)
• Open source

Monkey's Audio

• Symetrical codec (decodes at same speed of encode)
• An older and well known codec. No hardware support due to difficulty of decode
• 5 different compression modes, each with a different encode and decode time (fast (-c1000), normal (-c2000), high (-c3000), extra high (-c4000), and insane (-c5000)
• Compression ranges from moderate in speed to extemely slow and good to extremely good in level, decompression ranges from moderate in speed to extremely slow
• Uses APEv2 tags
• Error resiliency is questionable
• Has best support on Windows, and limited support on other platforms. For the most part a Windows codec
• No longer actively developed
• Formerly closed source, but currently open source

There isn't a wrong choice for lossless format as transcoding from one to the other is perfectly safe and extremely easy to do while keeping tags. Say for example you encode your entire collection one day to Monkey's Audio and you want to have support for a new device of yours with FLAC. You simply batch transcode your collection overnight and you'll wake up to a collection of FLAC files ready to go. Lossless to lossless transcoding is always lossless so you can't go wrong. The only downside to lossless compression of course is filesize, but if you have the space for it it's worth it as you'll never have a reason to rip your CDs again.

Lossy codecs

MP3 - MPEG-1 Layer III

• Most prolific of all current lossy encoders with hardware support almost everywhere
• MPEG standard with specifications open to those who are willing to pay for them
• Both open source and closed source implementations
• Has an extremely high quality implementation, LAME, which is open source and free
• Not natively gapless, however files encoded with LAME can be played back gaplessly by supporting players
• Old codec, finalised in 1992 (MPEG-2 variant in 1994)
• Achieves perceptual transparency at slightly higher bitrates than more modern codecs
• Doesn't scale down to low bitrates very well
• 320kbps is largest frame size
• Fixed frame sizes of 32, 40, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, and 320 for MPEG-1 and 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, and 160 for MPEG-2/2.5
• Supports sample rates of 32, 44.1, and 48kHz for MPEG-1, 16, 22.05, and 24kHz for MPEG-2, and 8, 11.025, and 12kHz for MPEG-2.5
• Hybrid subband/transform codec
• Slow seeking
• LAME actively developed
• Easy to decode resulting in increased battery life
• Uses ID3v1 and ID3v2 tags or APEv2 tags

Ogg Vorbis

• Not as prolific as MP3, however hardware support is gaining. Especially with Asian manufacturers
• Open standard with free specifications
• Open source
• Has an extremely high quality implementation, aoTuV, which is open source and free. Modern official Xiph versions are based off aoTuV b2
• Natively gapless
• Modern codec with its version 1.0 in 2002
• Achieves perceptual transparency extremely competitively with other codecs
• Scales down to low bitrates very well
• No maximum frame size
• Completely variable frame sizes
• Supports many sample rates
• Transform codec
• Fast seeking due to the container, Ogg
• aoTuV actively developed
• Complex to decode resulting in somewhat decreased battery life
• Uses Vorbis comments

MPEG-4 AAC

• Not as prolific as MP3, however has hardware support and a large corporation backing it: Apple
• MPEG standard with specifications open to those who are willing to pay for them
• Both open source and closed source implementations
• Has too extemely high quality implementations, iTunes and Nero. iTunes is free, but Nero is pay software
• Not natively gapless, however files encoded with Nero can be played back gaplessly by supporting players
• Modern codec, finalised in 1997
• Achieves perceptual transparency extremely competitively with other codecs
• Scales down to low bitrates very well
• Very high maximum frame size
• Completely variable frame sizes
• Supports many sample rates
• Transform codec
• Fast seeking due to the container, MP4
• iTunes and Nero actively developed
• Complex to decode resulting in somewhat decreased battery life
• Uses MP4 tags

Musepack

• Not very prolific at all, and hardware support is currently only limited to some Rockbox enabled DAPs (iRiver H1xx and H3xx)
• Open standard with free specifications
• Formerly closed source, but currently open source
• Implementation is of extremely high quality
• Natively gapless
• Hybrid old-new codec, based off MPEG layer II (MP2), but considerably more advanced
• Achieves perceptual transparency extremely competitively with other codecs
• Doesn't scale down to low bitrates very well
• Maximum frame size around 1300kbps
• Completely variable frame sizes
• Supports sample rates of 32, 37.8, 44.1, and 48kHz
• Subband codec
• Slow seeking due to flaws in the stream container
• Currently not very actively developed
• Easy to decode resulting in increased battery life (provided proper optimisations)
• Uses APEv2 tags

Choosing a lossy codec is based on more factors than a lossless one. If portabitlity and compatiblity with everything is most important go MP3. If you want to squeeze more bits out of your encodes or plan on going extremely small with them Vorbis (or possibly AAC) would be a better choice. I wouldn't recommend Musepack anymore because its benefits have been most overshadowed by Vorbis in particular.

I'd actually leave this between MP3 or Vorbis to be honest, but I can't make the decision for you. I can only guide you by presenting everything in an easy to read format.

Easy to follow Exact Audio Copy ripping guide
Easy to follow transcoding guide via foobar2000

Recommended encoders and settings

FLAC 1.1.2 (setting up to user, -5 is default)
WavPack 4.31 (setting up to user, normal is default)
Monkey's Audio 3.99 (setting up to user, -c2000 (normal) is default)
LAME 3.97b2 -V2 --vbr-new
Ogg Vorbis aoTuV b4.51 -q6 or Lancer based off of it (Select appropriate version for your CPU)
Nero AAC encoder Transparent profile
iTunes 6 192kbps VBR
Musepack 1.15v --standard