Mixing/mastering for mp3

D

Dropped Z

New member
With the prevalence of mp3 distribution amongst the internet these days, has much effort been given to mixing or mastering specifically for digital formatting? In other words, have people considered having two different mixes: one for your physical, WAV, or FLAC formats, and one for lower quality mp3 formats?

Because of the loss of fidelity on the mp3 conversion and the likelihood of them being listened to on crappy speakers, maybe a second mix should be made for that purpose. If so, how would you mix or master them differently?
 
AlanSwenson said:
One thing I do is make sure to record in 44.1 16 bit so that It gets resampled the least amount possible.
Click to expand...

I don't think 16bit is a wise choice. MP3 conversion isn't about resampling, but more about compressing data (not to be confused with audio compression), by discarding information that it deems "unnecessary" or less necessary, by taking advantage of audio masking. However, when you listen carefully, there is always a difference, although with 320k MP3 files the difference is almost inaudible compared to the source files.

Nevertheless, I'd still give it the best quality to work with.
 
AlanSwenson said:
One thing I do is make sure to record in 44.1 16 bit so that It gets resampled the least amount possible.
Click to expand...

Recording in 44.1 makes sense.

Recording 16 bit doesn't.



24 bit converted to 16 bit does not involve any resampling.

To over-simplify: Going from 48 to 44.1 means the computer has to throw out every single sample in the song and calculate new ones. Going from 24 bit to 16 bit every sample stays in the exact same place with the exact same amplitude. The only difference is the number that describes the amplitude doesn't have as many digits anymore (horribly simplified, I know).
 
This is gonna be a good thread!:D

I read a study recently comparing college students preference for music formats and audio quality. Now the sad part, most of the students surveyed prefered mp3 sound to wav or similar. They said it sounded louder.

Well, like the soccer moms say, the children are our future!
 
As far as how to optimize a mix/master for mp3, it seems the two most effective things directly oppose each other.

First, if you assume mp3s are going to be played most frequently on bad ear buds and computer speakers and almost never on a $2000 stereo, then I'd say limit the dynamic range a bit (note, current day dynamic range limiting is already more than this would need). Bad playback systems have a small dynamic range to begin with so people probably wouldn't notice the damage. I can not stress enough that I don't mean to limit any where near current loudness levels. That passes the point where it is assisting small speakers.

Second if you look at how the encoders work, it seems an mp3 would encode cleaner with more dynamic range since they rely heavily on loud short transients masking the material imediatly after the transient. Limited dynamic ranges give the encoder one less thing to cut, so data that is more obvious might get the knife instead.


Point one directly opposes point two.


Still, since I think current loudness is far beyond what even the weakest speaker needs to be audible I say the best thing you could do to mix/master for mp3 would be to leave a larger dynamic range than the current standard with replay gain data included.

In fact, you could leave the micro dynamics in tact (don't shave off the tops of drum hits with a limiter) but reduce the macro dynamics in the mix (just don't make the verse so soft).

That way you've got full dynamic range so the encoder can take full advantage of the transients' masking properties, and people could still listen on small speakers because sections of the song don't go quiet.

The replay gain data you include keeps your song in line with other songs in a playlist (if they also have the replay gain data), so it turns out all good.




Also something could be done to optimize the joint stereo feature. I'll guess and say LCR mixing probably encodes better in joint stereo compared to panning all over the field, but I don't know.
 
Last edited:
I'm confused...is this a mixing or mastering question...is it in the right forum? ;)

I know some guys like to adjust their audio processing for a specific medium...and I guess in some cases, that may be necessary.
But for overall music production, I would still aim for a Hi-Q medium and let the MP3 crap fall out as it may.

The notion of even considering how to do a "great mix" for earbuds coming off an iPod...seems rather oxymoronic to me.

AFA EQ/range...the smaller/crappier playback devices tend to roll off stuff on their own anyway....so why do it during production?
Not to mention...should someone put up your MP3/earbud/iPod mix on a different and/or Hi-Q system, it's going to sound like ass.

I still go by the mentality that if you mix it for Hi-Q...it tends to translate well on other stuff...give or take the shortfalls of the particular playback device.
And personally...the college MP3 crowed is probably NOT going to care all that much of you "optimized" it for their crappy MP3/earbud/iPod players. :D
 
I mix exactly the same 'for MP3' than I do normally... MP3 compression is a lot better than most give it credit for (most people completely knock it and say it sounds horribly different without doing any kind of proper blind testing) and really won't cause any audible changes that make it worth doing a different mix, let alone any perceivable differences at all (with higher bitrates). Even with lower bitrates, a good mix should translate "well" however badly its mangled and mushed up. I see too many people in the MP3 clinic blaming a poor mix on the fact its 'as a crappy MP3'... wake up, its a 192kbps MP3 you've posted... if it sounds crap like that, its gonna sound crap anyway!

On the other hand, the exception to this is Myspace. I often do separate mixes for songs destined for Myspace that are mixed quite bright with weak low-end, have some low mids notched out, and a wide high-end boost (most of which is lopped off by the LPF of the encoder, but means that what remains is crisper). Just with these changes alone, once Myspace has had its what with it the sound ends up sounding more like an 'ok' encoded 96kbps MP3, rather than a poorly encoded 64kbps.
 
Myspace is an interesting situation. In that case you know it is going to be on computer speakers. I'm going to guess that they use joint stereo encoding... but I couldn't say for sure. If they did, it might be wise to mix mono for Myspace only. It could be a huge improvement.
 
Chibi, that is an interesting thought as myspace is a really large promotion medium. Almost everyone is on myspace and it shouldn't be ignored when you try to distribute your music, so the track should sound as good as it gets within the limited bandwith.

Mastering for mp3 should essentially be treated different just like mastering for CD, SACD or vinyl is treated different. Mastering engineers always had to take the output format into account. A transfer to vinyl needs different treatment as a transfer to tape. Same goes for CD, SACD, DVD and last but not least mp3. Now, the difference between CD and mp3 might not be as huge as between CD and vinyl, but it is noticable and since the encode/decode process alters the waveform (it is completely recalculated on playback), tweaking might become very important. A hot tune reduced to 128kbit mp3 might clip quite a lot when played back on consumer gear although the audio source never exceeded 0dBFS. As I already wrote in another thread, the following papers are a good - albeit techy - read regarding digital mastering, different output formats and intersample peaks.

Cheers
Tim
 
Thanks for posting this thread. I am interested in the topic and not all too well informed on the subject matter. I will check out the links provided to get some more information. Coming from a less than informed position I have defaulted to using the same mix that is intended for general user chosen release.

My recurrent problem is the artist's insistence on a "rough mix" to listen to improve performance, make notes etc. in planning for the next session. Son of B&%h, up it goes on Myspace! (at least my name isn't associated with the damn thing). Sorry for the digression,the aforementioned is my problem not related to this thread. I am interested in the net effect of Myspace on an a song and information on encoding schemes. On the songs I have worked on and then later listened to from Myspace the low end sounds weak. Be well all and thanks for the links. Anymore tech articles, I would love to read them as they are a radical departure form my day gig and I am always looking to learn more "current" information as I come from an older technology background and am just now playing catch up. Catsup. :) One final thing. 44.1 = sampling rate, 16 bit and 24 bit not sampling rates.
 
mattr said:
On the other hand, the exception to this is Myspace.
Click to expand...

I've not done anything different for MySpace.
Sure, it doesn't sound as good as my CD cuts...but the overall "degrading" that occurs, hasn't hit any one or two particular areas of the frequency range...it's just compress (and mess up) all of it about the same.
At least that's been the case with my mixes.

Same goes for MP3s done at 128 and those done at 160 or above.
Each one does it's part to mess up the sound at different degrees, but I've not seen crazy tonal differences from one to the other where I thought I needed to mix/EQ specifically for MySpace or whatever, and once you get up past 200...even the MP3s sound decent, though still not CD-quality *if you play back on a decent system*....if you play back on a crappy system, it doesn't matter as much since the system "equalizes" the quality.
I still feel a quality mix done on a quality system will always translate the best across most system. You may be able to “tweak” a little for a particular format/system…but IMO, you can get caught up chasing ghosts like that…and you really have NO idea what the listener is doing.
I figure...go for best quality you can get...and let the listener ruin the quality any way he/she sees fit. ;)
 
miroslav said:
I've not done anything different for MySpace.
Sure, it doesn't sound as good as my CD cuts...but the overall "degrading" that occurs, hasn't hit any one or two particular areas of the frequency range...it's just compress (and mess up) all of it about the same.
At least that's been the case with my mixes.

Same goes for MP3s done at 128 and those done at 160 or above.
Each one does it's part to mess up the sound at different degrees, but I've not seen crazy tonal differences from one to the other where I thought I needed to mix/EQ specifically for MySpace or whatever
Click to expand...
It's not a matter of EQ. I agree that won't do anything. It's a matter of playing towards the perceptual encoder.

Uncrushed peaks encode better because they mask better in the time domain.

Mono encodes better because half as many channels means twice the bitrate on the channel that remains.

LCR mixing would probably encode better in joint stereo (especially if stereo mics/effects were collapsed to mono) because the math would be so much more straightforward.
 
Chibi Nappa said:
Uncrushed peaks encode better because they mask better in the time domain.
Click to expand...

Yeah...I tend not to crush...maybe that's why it's translating OK.

AFA mono...I just let that fall where it will.
I mean, I check mono when mixing just to see how it will sound if it has to be heard in mono, but I live in a stereo world, and that's my preference for mixing, so if some MP3 website does mono (I think CD Baby does that too)...so be it.
I have my stuff on CD Baby, and I really wasn't disgusted by their MP3 encoding of my songs.

To be honest...I use to be pissed about having to even do MP3s...and then I decided to just plain ignore them!
I mean...they get used because they are a "necessary evil"...but I personally ignore MP3s as a prefered listening format or my mixing concern. *shrug*
 
Great discussion - wish I had been around today to participate.

What I'm taking from this so far is (keep in mind a lot of the discussion was over my head, technically):

1) Mix and master so that it will sound its best listened to through a quality stereo system, and let mp3's do what they will do.

2) A well encoded mp3 sounds pretty damn good unless you have it on a system good enough to really notice the differences AND you know the original well enough to compare


The myspace thing is a good point, or any other site where your music is previewed. It might not be a bad idea to mix "hotter" strictly for the files that will be streamed on MySpace, your website, etc. I'm not convinced that the return is worth the effort though.
 
Hey Z,

Dropped Z said:
Great discussion - wish I had been around today to participate.

What I'm taking from this so far is (keep in mind a lot of the discussion was over my head, technically):

1) Mix and master so that it will sound its best listened to through a quality stereo system, and let mp3's do what they will do.
Click to expand...

Not necessarily. If you tend to "crush the peaks", meaning trying to reach the maximum peak on your mains without actually clipping, CDs will start to distort on low end consumer gear due to the crappieness of the digital-to-analog converters. Have a look at this short article where intersample peaks are explained in easy terms. Distortion with mp3 is even worse with myspace being a lot worse than a decent mp3 so the output will sound even more harsh and overcompressed and partly distorted on cheap consumer gear. Most consumers own cheap gear and most computer systems have crappy speakes attached (or built in) so a safe bet would be to keep that in mind and master for cheap gear and not for professional gear.

Dropped Z said:
2) A well encoded mp3 sounds pretty damn good unless you have it on a system good enough to really notice the differences AND you know the original well enough to compare
Click to expand...

That's right. well encoded mp3s can sound quite good. That's if you stay about 1 to 3 dB below clipping. It depends. A HOT HOT HOT mix can sound good as an mp3, too, if you have a decent system to play it on with good digital-to-analog converters with a lot of headroom (professional audio interfaces) but it will sound crappy on most consumer gear and that might include standalone mp3 players.

Dropped Z said:
The myspace thing is a good point, or any other site where your music is previewed. It might not be a bad idea to master "hotter" strictly for the files that will be streamed on MySpace, your website, etc. I'm not convinced that the return is worth the effort though.
Click to expand...

Actually the opposite is true. It would make sense to mix with more dynamics (should do that anyway, I have to remind myself every time ;)), cut back on the compression, stay way below 0dBFS. I don't know if I would do a completely different mix but I'd tweak at least the main bus volume to stay below -1dB and cut back on main bus compression.

Tools like the SSL X-ISM meter show you intersample peaks and that helps with transfer to WAV/CD but with mp3 the waveform is calculated on decoding, that means when the consumer plays your song. So the only method to check if the mp3 clips would be to decode the mp3 yourself, import it into your DAW and check it for peaks/intersample peaks. A source that's completely non peaking can actually introduce regular overs (peaks) when ecoded and decoded again. So you can at least check for your own decoder, that the mp3 in question is not distorting on playback. I don't know if decoders are all the same but I'd think they're not so you don't know if it will clip on another system, even if it was good on your system. The best bet really is to stay at least below -1dB. If you want to be really safe, stay below -3dB. That's what Lund suggests in one of the papers I linked.

Cheers
Tim
 
Last edited:
You must log in or register to reply here.
Back
Top