tigersoul Posted July 1, 2012 Share Posted July 1, 2012 Hi folks! Some of you may not know it, but I assume most of you do. When you bounce to mp3 from logic and it's a mastered or otherwise relatively loud track with limiting, the resulting mp3 will CLIP. You have to back down the out ceiling of the final limiter to about -1db to stay clear of the ceiling in the resulting mp3 file. It's also good practice, to most, to leave 0.3db or so headroom for poor DA's and Intersample Overload, meaning a total of about -1.3db is needed. This is quite alot of you ask me. If I don't do any of this and just export at 0db it still sounds fine on my system despite the mp3 clipping but I've had reports from other users using lesser quality speakers and interfaces that they hear distortion. What is you practice here regarding this? Do you look in the final mp3 file for that -0.3db headroom, do you care at all about the mp3 clipping, what's your take on this? Link to comment Share on other sites More sharing options...
BJG Posted July 1, 2012 Share Posted July 1, 2012 Hmm.. I really only use the mp3 bounce interim while working on project revisions to avoid a gazillion 30MB+ bounces. Clipping etc etc are really not of concern. Any final mp3 delivery, just use a real LAME encoder. -B Link to comment Share on other sites More sharing options...
lagerfeldt Posted July 1, 2012 Share Posted July 1, 2012 Hi folks! Some of you may not know it, but I assume most of you do. When you bounce to mp3 from logic and it's a mastered or otherwise relatively loud track with limiting, the resulting mp3 will CLIP. You have to back down the out ceiling of the final limiter to about -1db to stay clear of the ceiling in the resulting mp3 file. This is due to inter-sample peaks above full scale. The amount of dB depends entirely on the nature of the ISPs. It's also good practice, to most, to leave 0.3db or so headroom for poor DA's and Intersample Overload, meaning a total of about -1.3db is needed. 0.3 dB is a somewhat arbitrary number and isn' related to actual ISPs. Consider the 0.3, 0.2 or 0.1 dB ceiling as an additional buffer, but it won't be enough to catch ISPs in heavily clipped or limited material at all. This is quite alot of you ask me. If I don't do any of this and just export at 0db it still sounds fine on my system despite the mp3 clipping but I've had reports from other users using lesser quality speakers and interfaces that they hear distortion. Try playing back on a laptop or iMac. Assuming there are many and loud ISP's you will notice audible pumping as the system protection limiter kicks in. On Spotify it will cause pumping as the always-on limiter kicks in. On some converters it will cause audible distortion, on others it is less noticeable. On iTunes it will cause the AAC conversion process to lower the resulting file, since it takes places in floating point with peak normalizing down to 0 dBFS in case of 0 dBFS+ signals. I.e. a very loud, clipped master could be lowered a 1 - 1.5 dB once it's available in the iTunes store. This is hardly ever discussed, but yet another reason to avoid extremely loud masters. What is you practice here regarding this? Do you look in the final mp3 file for that -0.3db headroom, do you care at all about the mp3 clipping, what's your take on this? There are several ways of avoiding 0 dBFS+ signals: A) One is to lower the output level until there are no more ISPs above full scale. Use an inter-sample peak meter to confirm this. If you want to be on the safe side then add a bit more headroom. You can use clipping and hard limiting, but especially clipping will be self-defeating as you need to lower the output ceiling to compensate. B) Use a 4x oversampled limiter (the limiter should upsample the side-chain only) at the very end of the chain to catch ISPs. If you want to be on the safe side then check with an ISP meter and add a headroom buffer of a couple of fractions of dB (hence the often referred to 0.3 dB, which really is a fairly arbitrary number). Using an oversampled limiter is my preferred method as it allows you to apply all types of processing, including intentional clipping at various stages, but it also exposes the paradox of using clipping or hard limiting since the oversampled limiter will be pushed harder. The solution is to have the oversampled limiter on while performing the earlier steps since thats the only way of evaluating the actual sound and not be surprised later. C) For AAC you need to be even more conservative and preferably use Apples own measuring tools. They're freely available as part of the Mastered for iTunes programme. For actual MFI files I recommend you use a EBU R128 approved True Peak meter and aim for the standard -1dBTP. This is not mentioned in the MFI papers, but it will solve a lot of problems in one go. Link to comment Share on other sites More sharing options...
tigersoul Posted July 1, 2012 Author Share Posted July 1, 2012 Thank you, that was a very informative reply to say the least! I've tried a inter-sample peak meter to check my master, and it found it clipping at 0db, at -0.1db however the problem was gone, according to the meter. SSL X-ISM was the name of the plugin and this would be a great solution if MP3 compression wasn't part of the picture. It is however, and to get a FINAL headroom of -0.3db in the mp3 file, I need to dial in -1.3db in this particular project (which is pretty loud, but not VERY loud by any means). A few questions I'm not quite clear about: - In solution A you are using an IS peak meter, are you running the finished MP3 file in this meter to verify, or the project BEFORE bouncing to mp3? - Would a 4x oversampled limiter at the end of the chain fix the problems with clipping in the bounced MP3 files? So far I've not been able to find a method of knowing what the peaks in the resulting MP3 will be like. I've had to encode, check it and then adjust the limiter appropriately to reach the desired -0.3db headroom in the mp3 file. You also wrote that 0.3db isn't going to be enough to catch ISP's in loud limited material, how do we then know IF we are doing ok with our MP3 file or not? Can I run it, like I mentioned above, in the IS peak meter to verify? Link to comment Share on other sites More sharing options...
triplets Posted July 1, 2012 Share Posted July 1, 2012 Run the SSL xism plugin before you bounce. Best way in my opinion. Link to comment Share on other sites More sharing options...
tigersoul Posted July 1, 2012 Author Share Posted July 1, 2012 Run the SSL xism plugin before you bounce. Best way in my opinion. - Like I said above I have done this, it says all fine at -0.1db, and when I bounce I end up with an MP3 file that is pretty badly clipped. Link to comment Share on other sites More sharing options...
triplets Posted July 1, 2012 Share Posted July 1, 2012 You say all is fine. There shouldn't be any "white lights" on the SSL plugin during playback. If you see only one, that's an IS clip. Link to comment Share on other sites More sharing options...
tigersoul Posted July 1, 2012 Author Share Posted July 1, 2012 Was abit unclear, all fine = no white lights! Link to comment Share on other sites More sharing options...
triplets Posted July 1, 2012 Share Posted July 1, 2012 So that means that you're mix is way over-compressed. Louder doesn't mean better. Link to comment Share on other sites More sharing options...
tigersoul Posted July 1, 2012 Author Share Posted July 1, 2012 It's an 8 on the TT Dynamic Range meter, that's loud, but it's far from very loud. Either way, I'd prefer having control over this and understanding the problem. I don't like the loudness war much at all, but if my music is going to be played on crappy computer speakers for most of the time, competing with extremely limited tracks I don't have much of a choice as an emerging artist if I want people to feel it "sounds good". Personally, I'd rather just calm down the limiter, but I don't listen to music with speakers weighing less than a few eggs. Link to comment Share on other sites More sharing options...
triplets Posted July 1, 2012 Share Posted July 1, 2012 If you'd rather be loud and sound distorted than less loud and clean, that's your choice then. Link to comment Share on other sites More sharing options...
tigersoul Posted July 1, 2012 Author Share Posted July 1, 2012 I'm shooting for relatively loud but clear, hence this discussion Link to comment Share on other sites More sharing options...
triplets Posted July 1, 2012 Share Posted July 1, 2012 I know, and as you can see there's a sonic limit to loud and clear, and I think you have reached it. Link to comment Share on other sites More sharing options...
tigersoul Posted July 1, 2012 Author Share Posted July 1, 2012 Of course there's a limit, I was just interesting in how others handled this problem, how they measure for it and act around it. Also I think some people missed my focus here at the MP3 compression which easily can clip even from a -0.3db source if it's limited. I was wondering if there was any better approach than actually encode, check, and re-encode until you have what you want. Link to comment Share on other sites More sharing options...
shivermetimbers Posted July 1, 2012 Share Posted July 1, 2012 8) ... and I still have .2dB to play with. Even though the mix isn't absolutely perfect, the song is full and rich enough to play back on a laptop speaker or a thousand megawatt system without distortion. The mp3 format seems to add a slight bit of info to the final bounce. To top that off, whatever playback system used may add some more. A Clean mix is essential and there is a point where it ain't going to git any louder (mix wise). Putting the laptop volume at full and listening on the laptop speaks or thru headphone should put a smile on your face if you hear no distortion. This doesn't mean that you can skip the Professional Mastering process and compete with the Grammy winning artists, but chances are that you won't be in that league anyway. Link to comment Share on other sites More sharing options...
lagerfeldt Posted July 2, 2012 Share Posted July 2, 2012 I've tried a inter-sample peak meter to check my master, and it found it clipping at 0db, at -0.1db however the problem was gone, according to the meter. SSL X-ISM was the name of the plugin and this would be a great solution if MP3 compression wasn't part of the picture. It is however, and to get a FINAL headroom of -0.3db in the mp3 file, I need to dial in -1.3db in this particular project (which is pretty loud, but not VERY loud by any means). That sounds like something else is going on. Do you have 10 Hz filtering enabled during MP3 conversion? Disable that and try again. A 4x oversampled ISP meter should catch most ISP's, but you're welcome to send me the test file (WAV) and I'll check it out. Link to comment Share on other sites More sharing options...
tigersoul Posted July 2, 2012 Author Share Posted July 2, 2012 I can report for those interested that after some testing I can conclude that Logic's mp3 encoder is pretty poor when it comes to peak control. LAME is very controlled and outputs a file which is around +-0.1db of the source in regard to peaks. Logic's encoder however can spit out excess of a full db. I tried a meter for true peaks and found my project spat out +0.7db analog peaks (yikes) I tested using ozone at a friends studio to control this and still juice out good loudness and that worked very well. I also realized that my mix was a little TOO hot so I backed down abit on the limiter and now I'm pretty happy with result. I learned alot too. Thanks for the tip on the true peak meter, it really helped in understand what's going on. Link to comment Share on other sites More sharing options...
dhainlin Posted September 11, 2012 Share Posted September 11, 2012 Ok, I was struggling with mp3 bounces that were clipping (aif versions were maxed at -0.3db) and so I did some experimental bounces and found that mp3 needed to be bounced at almost -1.0 db max gain to prevent clipping. This was a fairly compressed/limited dance track. I was checking with Soundtrack Pro's analysis tools. In this case the mp3 still had peaks of -0.2 (meaning the mp3 encoding was overshooting by about 0.8db) I then noticed the 10 Hz comments in this thread and I removed it and now a -0.3dB peak mix doesn't clip (tho gets very close!). So... turning off the 10Hz filter seems to tame the ISP overshoot when bouncing MP3s. I also will set my limiter max gain to around -0.4 and will ALWAYS check for clipping of Logic bounced mp3s. Thanks for the pointers - was driving me crazy. Link to comment Share on other sites More sharing options...
tigersoul Posted September 11, 2012 Author Share Posted September 11, 2012 Personally I do now use a limiter with inter-sample awareness that outputs to -1.0db, this ensures I never ever get any problems with ISP's or poor soundcards and I never get reports of distortion anymore. I can't recommend this approach enough and you CAN lose 1db folks, it's not that bad, crank the limiter 1db more if needed, it'll produce virtually the same thing. Link to comment Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.