Mixing & Mastering For MP3 Release

Hi, I wanted to outline my plan here and see if it looks kosher to some of the wiser forum members around here. 

I'm mixing and mastering an album for digital download release. My plan is to master a mix to a ceiling of -3 db and export it as a 32-bit WAV.

Then I'm going to convert the -3 db WAV to a 320 kbps CBR MP3. Then after the conversion, I'm going to hard limit this file up 2.9 db to a -0.1 db ceiling (with very light if any limiting taking place) and save and ID tag the resulting file. This is the higher quality MP3 download.

Then I'm going to go back to the 32-bit WAV file and perform the same process, converting it to 128 kbps CBR MP3 and limiting it to -0.1 db. This is the lower quality MP3 download.

My questions are:

1. Is mastering to -3 db before conversion conservative enough to avoid clipping on the 320 kbps MP3 and the 128 kbps MP3?

2. Should I hard limit the converted MP3 files to -0.1 db like I'm planning? Or can I get away with going to 0 db (since no more conversion will be taking place)? Or should I be more conservative and hard limit these MP3s to -0.3 db?

Thanks so much for your reply!

Thanks for your reply, bitflipper. I was hoping you'd show up here. 


1. Thanks!


2. Very interesting. Is this because an MP3 player converts an MP3 file back to WAV as it plays and during this conversion has a chance of clipping if my MP3s are mastered too hot?


I suppose an MP3 release that is 3 db quieter than others is better than an MP3 release that is clipping.

Thanks for the replies! I've decided to make a master to -2 db, and then convert and use my ears to make a judgement on whether I should lower or raise the db level depending on how the converted MP3s sound. (I've got to remember to make music with my ears, not my eyes! )

bitflipper


In fact, your average iPod-damaged listener wouldn't even notice real clipping.

I see this as kind of like a cook spitting in the food. Sure, most people won't notice, but it's still gross.

Unfortunately, that's one of the things SONAR has always been missing. I use an external editor (Adobe Audition) to calculate amplitude statistics.

SPAN does have a display labeled "RMS", which the developer says is a cumulative value. Theoretically, that's what you'd want; you'd only have to play the portion of your song you're interested in and then look at the readout. Not as convenient as highlighting a section and having the software calculate it, like Audition does, though.

(BTW, IIRC the free editor Wavosaur has an average RMS feature. Audacity, I think, does not.)

However, the RMS display in SPAN does not appear to work as I think it should. It keeps climbing the longer you run it and always ends up reading much higher than the actual average RMS. I have asked Aleksey about this, but something's getting lost in translation. The feature is either broken or does something completely different than what I thought it does.

The good news is you don't really need an average RMS calculator to get in the ballpark. SPAN can still serve the purpose just fine, if you're not hung up on specific values.

First, insert SPAN as the last plugin in the fx bin on the master bus. Pull up SPAN's "Stereo Mastering" preset, then go down to the bottom of the display and click the "Metering" button until it reads either "K-12" or "K-14". I also recommend clicking on the Settings button and setting the integration time to 300 and the release time to 1000. Make sure the "density mode" option is checked.

[EDIT: I realized after writing this that the integration and release times are hard-coded to 600ms for the K-n modes, as that is part of the K-system specification.]

(I won't go into the K-system or how to calibrate your monitors for it. Go to Bob Katz's site for complete instructions and to download the calibration file.)

Now, as you play back your song, watch the vertical meters on the right-hand side. You want them hovering around 0 as an average. Adjust the threshold on your master limiter until you achieve this target. Don't worry if it jumps up to 4 or 6db into the "red", as long as on average it hovers around zero.

If your mix feels too squashed using this method on K-12, choose K-14 or K-20. Same method applies for all three.

This is a somewhat over-simplified process, but it'll get you in the ballpark if you're at a loss as to how loud to master your mixes. Most important, it will allow you to achieve consistency from one song to another.

Thanks very much, Donny!

I did come across that after investigating a little.  I usually do normalize for peak value and compress a little there, but had never messed with the RMS options...that was pretty neat.  I did get SPAN...that's pretty slick.

There's a lot of good information in this thread! Thanks everyone who contributed!

I ended up mastering to -1 db. Then I converted to a 320 kbps CBR MP3 and uploaded it to Soundcloud. I'm guessing it peaks at -.6 db after conversion. I don't hear any artifacts, and it sounds as good as my rusty mixing skills are capable of.

I think the *paradigm* is limit to 0.3 or even 0.2 dcbls for mp3 normalization.

A problem I have seems that some Bluetooth systems can't amplify enough waveforms loud enough, i.e., mp3's whose peaks are at -3 dcbls: Bluetooth phones (which rely on WMAs and MP3s) come to mind.  Their radio signal needs more loudness to drive the blue-tooth car stereos, else I turn up the Samsung phone to 7/7 output-level where the distortion is (to my ears). 

(I always play at 5/7 loudness for the bluetooth phone to radio into the car receiver.  I may be wrong.)

For this reason, I must gleefully convert my printed master (-0.2 to 0.3 dcbls) directly to MP3 or WMA, for better car bluetooth useage.

I'm just curious - what are the MP3's you're all talking about being used for? What is the end game for those?

+1 For the "Limousine" tune. I checked it out a few months ago and downloaded it to my pc at work. Listen to it a few times a week without fail. Amazing tune. Hope that's not an issue bitflipper. I downloaded it after I read your opinion about sharing music. So thanks for sharing that great tune.