What the heck!

[Rusty K]

I just mastered a file and converted to mp3. The master when analyzed was at -.49 db both channels. After conversion to mp3 it's clipping. I've had this kind of problem with sample conversion using AA from the git go but this is the first time for an mp3 conversion. What is going on with this?

Rusty K

[LogicDeLuxe]

As lossless codecs remove bits from the spectrum, the resulting wave can peak higher then the original. Higher quality encoding with higher bitrates can reduce this effect.
And of course, you should enable 32-bit-loading (which doesn't clip) for MP3, if you ever should need to use one as a source.

[Rusty K]

LogicDelux,

Could you explain more about the 32bit loading? I'm not getting that. You mean keep it at 32 bit/24 for encoding? My mastering softward doesn't recognize 32 bit files. I guess I could master without dithering....confusing.

I tried the mp3Pro encoding but I got the same clipping of the mastered/dithered file.



Thanks,
Rusty K

[lpdeluxe]

The actual processing is compatible with 24-bit, and you can use that setting on other gear (it's what I do). 32-bit has been explained to me, but all I know is that it works with 24-bit.

It shouldn't cause any issues.

As for the clipping, damfino.

[Rusty K]

lpdeluxe,


Hey man great to see a familiar face/text here. How's it going? Man you should see my new 64 reissue Fender Jazz Custom Shop. I'd been playing an American Fender Zone (active) but with my Ashdown amp all the electronics seemed like over kill. I've made the transition to my new toy with no problems.

On the subject....I guess then I will have to do some experimentation on my own today. I hadn't thought of it but I'm wondering if this issue has anything to do with post dither processing when I convert a master to mp3. I already knew the stuff about 24bit/32bit and AA. I use TRacks to master and it can handle the 24bit just not the 32 but I can master without dithering and covert the file to mp3 at that in between stage of the process. Or I can just master at say -1db but it's disappointing that without knowing what the final mp3 db level will be I'm unable to run it up close to 0db. That just doesn't seem right to me. There has to be a logical explanation either that or it's in the AA processing.

Take it easy bro...
Rusty K

[LogicDeLuxe]

Originally Posted by Rusty K

Could you explain more about the 32bit loading?

Ok, it's a but confusing, since that particular option is in the save-dialog in the advanced mp3-options called "Set all decoding to 32-bit". After you set this option, you can load any mp3 file, and it doesn't clip anymore. If it's too hot, you can just normalize or limit it.

You mean keep it at 32 bit/24 for encoding?

That's a different story. But, yes, there is absolutely no need to dither to 16 bit when it's for mp3 encoding.

My mastering softward doesn't recognize 32 bit files.

Then use 24 bit. LAME supports it.

I tried the mp3Pro encoding but I got the same clipping of the mastered/dithered file.

mp3Pro is another completely different matter. There is much less hardware support for it than regular mp3. And it depends much on the style of music how good it works.

Originally Posted by Rusty K

Or I can just master at say -1db but it's disappointing that without knowing what the final mp3 db level will be I'm unable to run it up close to 0db. That just doesn't seem right to me. There has to be a logical explanation either that or it's in the AA processing.

You can't expect the peaks to be the same as in the source wave, this is just impossible. It's in the nature of altering the wave of any lossy codec. Mastering at about -1 dB is actually a good idea to begin with. Then check where the peak really is in the encoded file, go back to the master and boost it by that value. Then you should peak at exactly 0 dB.

[Rusty K]

LogicDeluxe,

Thanks man....If I may take just a bit more of your time let me clear up a couple of things to make sure I'm straight on it.

So under "save as" mp3 "options" tab I click "advanced" I check the box "encoding set to 32bit"? Did that and the file is even hotter.....so it doesn't matter that I'm processing a signal that I've already dithered?

I'm leaning toward mixing to -1db and following your last instructions.

I don't really understand how you can be taking so much data out of the file yet still be increasing the output level.

And yes I take my mixes out of AA as 24bit to master.

Thanks again,
Rusty K

[Rusty K]

Ok I've remastered to -1db and left the file 24bit with no dither. I encoded it normally (no 32bit) and it peaked at 0db in a couple of places but "clip restoration" says it's ok with 0% clipped.

I still have questions on all this though. Like what do the pros do about these issues when setting up mp3 files for download purchases?

Rusty K

[lpdeluxe]

I have gone completely to Fender. Now have a '51 P RI, a Classic '50s P w Seymour Duncan QP and TI Flats, a fretless J and a new P FSR Standard in natural ash, which is soon to get a fretless unlined maple neck. Life is good.

But I don't know why your mp3s are clipping!

[dobro]

Originally Posted by Rusty K

I just mastered a file and converted to mp3. The master when analyzed was at -.49 db both channels. After conversion to mp3 it's clipping. I've had this kind of problem with sample conversion using AA from the git go but this is the first time for an mp3 conversion. What is going on with this?

I'm thinking it's got to do with how maxed out you are on the overall level before you converted to mp3. I've had this happen too - the wav file mixdown is under the clipline, but the mp3 clips. You're what? -.49 dB on your overall level? That's *really* close to clipping. And that's what your meters are telling you, right? You can't trust meters to be more than approximate, I've found.

Try this: do a mixdown with the hottest signal maxing at - 2 dB. Do the mp3 conversion on it. Does it clip?

[LogicDeLuxe]

Originally Posted by Rusty K

So under "save as" mp3 "options" tab I click "advanced" I check the box "encoding set to 32bit"?

It's "decoding", not "encoding", taht's why I said, it's a bit confusing that this option is with the "save"-options when in fact it influences the loading only.

Did that and the file is even hotter

Not surprising, as you did clip at 0 dB before, and now, your peaks don't clip, but actually go over 0 dB. That's why I said, you can normalize or limit that thing, in case you really have to take an mp3 as a source file for some reason.

so it doesn't matter that I'm processing a signal that I've already dithered?

If you have to save it to 24 bit because your other software can't load float, use dither. You just shouldn't convert to 16 bit when it is for mp3 encoding, though.

I don't really understand how you can be taking so much data out of the file yet still be increasing the output level.

That's in the nature of wave form harmonics.
Maybe this example helps you understand: Generate a new wave of 44100 Hz, use "Generate" -> "Tones..." and create a 440 Hz squarewave at -9 dB (a few seconds will do). Then, use "Effects" -> "Filters" -> "Quick Filter..." use the "Flat"-preset and take the 344 Hz slider all the way down. See what happens with a mere omission of some frequencies? All the lossy codes are based on such omissions.

Originally Posted by Rusty K

I still have questions on all this though. Like what do the pros do about these issues when setting up mp3 files for download purchases?

I'm afraid, most don't even care. They probably just encode from the CD-master. I heared paid mp3's with 192 kbps having much more apparent artifacts than the average 128 kbps mp3 given the encoder isn't hopelessly dated.

[Rusty K]

LogicDeluxe

It's "decoding", not "encoding", taht's why I said, it's a bit confusing that this option is with the "save"-options when in fact it influences the loading only

Ok so checking this box insures that the file is read as 32bit before encoding?

Not surprising, as you did clip at 0 dB before, and now, your peaks don't clip, but actually go over 0 dB. That's why I said, you can normalize or limit that thing, in case you really have to take an mp3 as a source file for some reason.

Don't even want to go there. I want to avoid this situation all together.

If you have to save it to 24 bit because your other software can't load float, use dither. You just shouldn't convert to 16 bit when it is for mp3 encoding, though

This may be the most valuable new piece of info you've set me straight about. If I'm getting this right you're saying that each time I down sample I should dither not just at the end when I down sample for 16bit CD? I'm speaking strictly of wav files here. I didn't realize that I needed to dither when saving wav files at 24bit. I thought they were 24bit and AA simply read/interpreted them as 32bit float.

Generate a new wave of 44100 Hz, use "Generate" -> "Tones..." and create a 440 Hz squarewave at -9 dB (a few seconds will do). Then, use "Effects" -> "Filters" -> "Quick Filter..." use the "Flat"-preset and take the 344 Hz slider all the way down. See what happens with a mere omission of some frequencies? All the lossy codes are based on such omissions.

I'll try this experiment. I do understand also that various software and other media will show minute differences in levels of the same file. Some of my friends that I've been asking this question of say that I shouldn't worry about a little clipping, if I'm not hearing it. Maybe I should lighten up on this a bit.

Thanks so much,

Rusty K

[Rusty K]

lpdeluxe,

Nice guitars man! Years back I played a Dan Armstrong clear vinyl fretless. Do you remember that one? It was actually a pretty nice bass....pickup was hot for the time. I was in Guitar Center this week and they had a Rail or some equivalent hooked up to a Bose System with their new Tone Lock processing. It sounded great. I have always stayed away from uprights mainly because my hands are pretty small and I was afraid it would play me instead of the other way around. I've always wondered about the Rail though.

What kind of music do you use the fretless for?

Rusty K

[LogicDeLuxe]

Originally Posted by Rusty K

Don't even want to go there. I want to avoid this situation all together.

Then just check how far it goes over 0 dB, go back to the source file you encoded, lower it by that amount and reencode. You should peak at exactly 0 dB now.

This may be the most valuable new piece of info you've set me straight about. If I'm getting this right you're saying that each time I down sample I should dither not just at the end when I down sample for 16bit CD?

Dithering generally improves the resolution in mid and low frequency range, thus do it. The difference is by no means that drastic as with 16 bit of course. If possible, you should avoid dithering more than once, except you are dithering at a lower resolution each time, for instance first dither at 24 bit and 2nd dither at 16 bit.

I thought they were 24bit and AA simply read/interpreted them as 32bit float.

That's how AA works with any file format having a resolution better than 16 bit, since it is the best way to work with it.

Some of my friends that I've been asking this question of say that I shouldn't worry about a little clipping, if I'm not hearing it.

Depends on the style and instruments. The less harmonics the clipping instrument has, the more distracting it will probably sound. Also a clipping e-guitar surely sounds less disturbing than a clipping singer. Clips of 3 or less samples aren't usually much to be worried about, and most software don't even detect them as clips. Clips of 4 to 10 samples produce squarewaves in the range where the human perception is rather sensitive. And even longer clips really shouldn't be there as they pretty much replace the entire spectrum with sqarewaves with all their harmonics for serious periods of time.

Though, clipping mp3 aren't that much of a problem than clipping CD's for several reasons:
- Better mp3 decoders have a limiter or a volume slider you can turn down, thus preventing the clipping in the first place.
- The volume of mp3's can be changed after encoding in 1.5dB steps. Try mp3gain. It not only can normalize mp3's, it also can take an entire collection and match the percept volume using the replaygain standard which AA also deploys in its "Group Waveform Normalize..."-function.

[lpdeluxe]

I practice on the fretless. It sounds great on ballads, where you can swoop into a note for an accent. After playing Dobro for mumblety-'leven years, the intonation's not a problem.

I took it to a gig and played it for a few songs in the second set. After the set, the singer took me aside and said "That fretless don't cut it." Then the lead player took my arm and said "If it ain't broke, don't try to fix it."

They were referring to the fact that I had finally found the bass sound they loved, which was the Classic '50s P with a Quarter Pound pickup and Thomastik-Infeld Jazz Flats. Anything less than that, and they were dissatisfied.

So I am building a fretless Precision -- just ordered an unlined maple/maple neck from Warmoth yesterday. It'll go on the natural ash P (which already has a QP in it, and for which I have a set of TI flats at the ready). When I'm done I'll get the swoops **I** want and the sound **THEY** want.

But that's not why your mp3s are distorting.

[Rusty K]

LogicDeluxe,

Thanks so much for you help. The links you've provided are fantastic.
I've downloaded Mp3Gain and the youtube tutorials I've saved to continue to look over.

Just a couple of points to summarize....

So there is no need to dither a 16 bit file coming out of AA but 24bit requires dither (actually 2 dithers if I go back to CD)?

With an mp3 there is no need to dither ever from any format?

Rusty K

[Rusty K]

lpdeluxe,

Post a link for me to listen to some of your stuff. Pvt. if you like.

It's nice that your mates are listening that close. Often times for many, the bass is an afterthought . It needs to be there but they don't analzye it that much. I first noticed this in acoustic situations without a drummer many players just can't listen close enough to the bass to lock onto the tempo. I guess it's just not definite and percusive enough for them......they're just not listening.....it's a pet gripe of mine.

My band is liking my change from active. My son who is also a musician and works for Guitar Center (nice) in St. Louis kept on me saying that my active was too "Vegas" for the roots music I've been playing. Everyone in the band plays vintage gear. Although my Zone was a very good guitar, he was right. I'm playing much better. It's true, there is magic in the right quality instrument.

Rusty K

[LogicDeLuxe]

Originally Posted by Rusty K

So there is no need to dither a 16 bit file coming out of AA but 24bit requires dither (actually 2 dithers if I go back to CD)?

With an mp3 there is no need to dither ever from any format?

Sorry, I don't get what you are asking now. Could you give some examples?

[Rusty K]

LogicDeluxe,

Ok....If I have a session that's at 16bit all the way through (read by AA as 32bit float) there is no need to dither when I create a mixdown?

But....If I have a session that is 24bit I do need to dither?

Rusty K

[LogicDeLuxe]

Originally Posted by Rusty K

LogicDeluxe,

Ok....If I have a session that's at 16bit all the way through (read by AA as 32bit float) there is no need to dither when I create a mixdown?

But....If I have a session that is 24bit I do need to dither?

Rusty K

It doesn't matter what bit depth your session files are. You do the mixdown to 32 bit float in any case, and if your other software needs 24 bit files for input, then you dither it to 24 bit.

[Rusty K]

Ok...got it!

Thanks,
Rusty K