Mixing for .mp3

[RadioMoo]

This might not be the "right" forum, but I respect the opinions of a lot of folks who post here.

I've been hearing a lot lately about creating separate mixes specifically for mp3s. Does anyone have any thought/suggestions?

[bob katz]

Originally Posted by RadioMoo

This might not be the "right" forum, but I respect the opinions of a lot of folks who post here.

I've been hearing a lot lately about creating separate mixes specifically for mp3s. Does anyone have any thought/suggestions?

If people mixed (and mastered) with more natural dynamic range, the way we did prior to about 1990 and the invention of severe brickwall limiters and clipping practices, there would be no need to make a separate mix for mp3. For that matter, the whole world would sound better if we went backwards. The problem is that lossy coders like mp3, AAC, and the like, all depend on having space in the frequency and dynamics spectrum to do their work. The denser and more distorted a mix or a master, the harder it is to make a good sounding mp3.

So, your mileage may vary, but the best solution is not to squash or distort so much and your mp3s will sound much better.

[Saudade]

I was just about to post some related questions when I saw this thread, pardon me if I'm hijacking...

My upcoming project is an exclusive digital (online) release. Probably, Itunes or Zune and the like.

Does it mean there is lesser need to have it professionally mastered? Is it "more possible" to get away with "self-mastering" when compared to a CD release?

Would asking a professional mastering engineer to convert my wav files to the compressed format yielf better results? Will mastering engineers approach my wav file premasters differently if I told them the destination format is an mp3?

Is there such thing as a "reference level" when mixing down to a compressed format (say if I am skipping mastering), since from my experience converting a wav file to mp3 wouldn't yield the same "perceived loudness".

Would the tips the Bob mentioned (mixing down to leave "headroom", less compression/limiting etc) apply equally to any compressed format? Or are there any specific tips one can learn about if the destination format is AAC or WMA?

What are the foolproof ways one can make lossily compressed audio sound......good

[RadioMoo]

Bump.

Not a hijack at all; these are my questions. Thanks for articulating them so well!

[Darius van H]

A good mix will sound good as an MP3......the higher the rate, the better...simple as that.

99.8% of consumers can't hear the difference between a 192kbs MP3 and 16bit44k........well, maybe they can subconciously feel it, but that's a different kettle of penguins.

[Nordenstam]

Different psychoacoustic coders are similar enough to master for, except for different fixed filtering of the top end. Given an open "good" sound to start with, the end result with 192 and above bit rates can easily sound better than a lot of the hilariously distorted CD's of today.

IMO, the very most offending noise comes from coding one of those louder-than-everything-else CD masters in lossy format! In that regard, if the mastering is done without special thoughts to the demands of the lossy codec, it may even sound worse after mastering. It's a fairly new media and most of it's problems and shortcoming are quite hard to grasp and work around. Shopping for special mp3 mastering have a potential to make a difference.


Most important is BK's advice, which was spot on, as usual. Severe limiting and clipping confuses the coder.

Am far away from being an expert, but have been pondering this subject and done some experiments for some time now. Watching the *real* peak level seems to be of high importance.

A potential source of problems are the inadequate and unfortunately ubiquitous sample peak level metering. It only tells how far the sample value goes on each of the sample dots, not how far the actual analogue waveform will go on the output of the digi->analogue converter. Pushing the sample points towards the ceiling will often create waveforms that shoots way above zero when they get converted to analogue. Using lossy codecs are sort of like converting it to analogue inside the computer. When the decoder outputs a normal wave bit stream, it synthesizes the waveform from the stored data. A signal level that may look alright on a sample peak meter can sound horrible if it clips the output of the decoder synthesis. An oversampled peak meter, like the RME cards have, helps to keep an eye on the potential problem. Some limiters takes care of this automatically, but watch out for too much limiting in general.

Since the decoders plotting new and different sample points than what was in the original file, the peak level will also change a little, even when using oversampled peak metering and limiting. Sometimes 0.5dB of headroom is enough, sometimes a full dB of silence on the top is needed to be safe from clipping in the decoder. Try coding and decoding and have a peek at the waveform again. If the peaks grows beyond zero, turn it down, and/or do something else to push levels if that's the quest.


Given the peak level constrains outlined above, outrageously loud mp3 masters are probably next to impossible without loads of more obvious bad effects than in a similar CD master. Yet, mp3 jukeboxes are an area where loudness war somewhat makes a bit of sense. Nobody wants to have the softest track in the collection. There's no denial that a very soft track may miss some attention it would otherwise get when people skip through 200 tracks to find the next catchy tune. It takes more work for me to get something to loudness level X or Y on lossy coding format than on a CD, while keeping things within guaranteed safe levels. Most of the usual loudness tricks makes things worse on lossy coders.

I try to make things dense but not distorted, softly pushing compression and often using parallel techniques to make it sound massive without limiting too much. Only lightly treating the waveform tops with processes that doesn't create 'hidden peaks' or high frequency content that may alias back into the audible register. These master are never going to be frigging loud, but they definitely does sounds better than simply coding a regular CD master.


Most consumers play mp3's on most befittingly poor audio systems. EQ'ing a bit like in many of the mid-eighties long play records, with very little bass, no sub and often a 15-16kHz limited top end, seems to work well for these crappy listening conditions.

Perhaps an idea would be to get two masters, one for low bit rates and laptop speakers, another for high bit rates and full range audio. Guess most inet consumers would love the option of an 'audiophile mp3'(hah!) for serious listening and a more trashed one for regular playback on shitty speakers.


.. and there's probably loads more to say that I forgot about now. Like coding from higher than 16 bit originals which helps too if the coder supports it.



Cheers,

Andreas Nordenstam