normalizing to 0.0 and -0.0 db

[jupiter8]

Correct me if i'm wrong but doesn't intersample peaks happen only after DA conversion so whether or not samples are normalized has bugger all to do with it.

They do, but the DA conversion is based around 0dbfs, and so if you normalise too high you are more likely to get InterSample peaks.

I was assuming he was talking about normalizing samples used in music production and not about normalizing masters.

]Whilst the normalisation to any value doesn't matter within the digital domain, any processing may potentially use 0dfbs as a reference point, and introduce artifacts.

But then that would be obvious wouldn't it ? That's like saying you shouldn't overdrive stuff. If you like what you hear then so be it.

I read this monster thread at GS how recording lower would make your tracks sound more analog and it was so full with BS it was absolutely nauseating. People were claiming IS peaks were happening all over the place. The most laughable aspect of the whole deal was that all of a sudden you couldn't trust your ears anymore but if you just follow the formula the computer would sound more like a tape recorder.
It's fine recommending people to try just to get rid of some habits but to say it's the recipe for great sound and you will ruin your recordings if you don't is just laughable.

[cosmicdawn]

I dither, but I don't see the point and can't hear the difference in a full mix.
But I don't want to be accused for lack of dithering. he he.

[koalaboy]

I was assuming he was talking about normalizing samples used in music production and not about normalizing masters.

There shouldn't really be a difference. There's no reason whatsoever to need samples at, or close to, 0dbfs anyway as they'll be processed further down the line.

I read this monster thread at GS how recording lower would make your tracks sound more analog and it was so full with BS it was absolutely nauseating. People were claiming IS peaks were happening all over the place. The most laughable aspect of the whole deal was that all of a sudden you couldn't trust your ears anymore but if you just follow the formula the computer would sound more like a tape recorder.
It's fine recommending people to try just to get rid of some habits but to say it's the recipe for great sound and you will ruin your recordings if you don't is just laughable.

Well, they (assuming the same thread) were actually saying that if you leave yourself enough headroom, you don't have to worry about ever having such problems - not that your tracks would sound 'more analog', but if that's what you want to believe.

I'd also take Mr Frindle's advice above just about any posters here regarding that sort of thing. YMMV of course, and if you'd like to provide your experience as being more substantial, feel free.

Some of us accept that just because you can't hear something when you monitor, it doesn't mean it isn't there, or that it won't become more obvious down the line. If you ensure you have headroom to avoid things like intersample peaks (and can easily check them in software) then you won't have to worry about the problem.

(That does the thread serious injustice, but it's most relevant to this thread)

It certainly holds as a basis, that if it sounds okay, then it is, for you. You may have amazing ears, in which case it'll be good for many people. That's no reason to dismiss best practices that would eliminate any possibility of problems which you may (or may not) hear.

[jupiter8]

I was assuming he was talking about normalizing samples used in music production and not about normalizing masters.

There shouldn't really be a difference. There's no reason whatsoever to need samples at, or close to, 0dbfs anyway as they'll be processed further down the line.

While i agree that normalizing is pretty pointless it doesn't hurt either. And there's a huge difference between a sample and a master since one is meant to be played as is thus risking intersample peaks while the other as you mentioned is due to further processing,hence intersample peaks isn't a problem.

I read this monster thread at GS how recording lower would make your tracks sound more analog and it was so full with BS it was absolutely nauseating. People were claiming IS peaks were happening all over the place. The most laughable aspect of the whole deal was that all of a sudden you couldn't trust your ears anymore but if you just follow the formula the computer would sound more like a tape recorder.
It's fine recommending people to try just to get rid of some habits but to say it's the recipe for great sound and you will ruin your recordings if you don't is just laughable.

Well, they (assuming the same thread) were actually saying that if you leave yourself enough headroom, you don't have to worry about ever having such problems - not that your tracks would sound 'more analog', but if that's what you want to believe.

http://www.gearslutz.com/board/so-much- ... tored.html
This one where he points out how incredibly important gainstaging is when in fact it is not. I believe later they claim that is peaks happens all over the place and when someone points out they don't they change the goalposts and mentions some plugins use 0 db as a reference yada yada yada as one is completely unable to make those kinds of decisions oneself.

I'd also take Mr Frindle's advice above just about any posters here regarding that sort of thing. YMMV of course, and if you'd like to provide your experience as being more substantial, feel free.

Now i didn't find any posts by Mr Frindle in that thread i'm vaguely familiar with his opinions. He's the intersample peak high priest is he not ?
I don't have anything against avoiding intersample peaks but to try and avoid them where they don't occur is just plain dumb.

[koalaboy]

Now i didn't find any posts by Mr Frindle in that thread i'm vaguely familiar with his opinions. He's the intersample peak high priest is he not ?

He doesn't start posting until page 3. Maybe if you don't notice them, they don't occur

He does explain why intersample peaks and dithering are important though, later on. Whether that makes him a high priest, I don't know. Perhaps his work at SSL and Sony Oxford help with that.

I don't have anything against avoiding intersample peaks but to try and avoid them where they don't occur is just plain dumb.

That's the point. You don't know where they occur unless you look, and they're far more likely if you normalise to 0dbfs.

Most people can't hear aliasing, but they want their plugins to stop doing it.

[jupiter8]

Now i didn't find any posts by Mr Frindle in that thread i'm vaguely familiar with his opinions. He's the intersample peak high priest is he not ?

He doesn't start posting until page 3. Maybe if you don't notice them, they don't occur

He does explain why intersample peaks and dithering are important though, later on. Whether that makes him a high priest, I don't know. Perhaps his work at SSL and Sony Oxford help with that.

Found him. Can't say i see anything he says that contradicts what i say. Seems we're in agreement. One note about authorities:you can be one and still be wrong. You may not take my advice over Roger Nichols since he recorded that album everyone loves but that doesn't make him right and me wrong when it comes to sample rate discussions.

I don't have anything against avoiding intersample peaks but to try and avoid them where they don't occur is just plain dumb.

That's the point. You don't know where they occur unless you look, and they're far more likely if you normalise to 0dbfs.

I know where they occur. At the DA conversion and nowhere else which means if i choose to normalize my samples for whatever reason i don't have to worry more or less about intersample peaks as one has nothing to do with the other.

People worry about so much insignificant shit these days it's laughable. Dither is one of them but there's plenty more. Anyone who says "You must always" or "You must never" is plain wrong. Simple as that. There are some good practices but if you don't know why you're doing them they're pointless.
Take normalizing for example: people say you should never normalize. Why ? What is so dangerous about normalizing ? Learn why it is mostly pointless and learn why it isn't such a bad idea after all and make up your own mind.

[koalaboy]

You should probably read more of the thread if you think you agree, especially regarding dither, but that's an aside. I'm sure he could be wrong, but I've not found any decent points towards that fact.

The main point is, that if you *know* all of the intricacies, you'll likely look out for them, and being aware of them is half the problem.

To just go around saying "normalise to 0dbfs is fine" is bad, because a lot of the people who hear the information won't have the slightest clue when they should or shouldn't care about clipping, intersample peaks (and likely not know they exist, even at the DAC) etc.

The standard answer seems to be 'aim for as close to 0db as possible' without context of production stage or usage.

If you don't worry about intersample peaks for samples, and normalise to 0dfbs, what about someone else who uses those samples and just masters to 0dbfs. Okay, so they should know what they're doing, but if your samples are safe in the first place, they'd have to go purposefully go over the threshold to break them.

You talk about generic advice being bad - but the OP was wanting to change -0.0 into 0.0 - That already demonstrated either complete lack of knowledge, or wanting very fine control of the sound - if the latter, then intersample peaks are potentially very important, and if the former.... well, none of this likely matters but at least they should be given all the information.

[flux82]

not too worried bout intersample clipping id just like them all to hit 0.0db but when loaded back in the daw for mixing you can always lower them again.. or even batch normalize them all to -1db..

but.. like i said i want to get them all to 0.0db.. that other audio editor doesn't do it either..

so im back were i started... tho i know it can be done.. i've did it before months ago no problem, would it be a vista/run as xp mode thing with applications? would this have something to do with the way its calculated.. or would it be something to do with different types of wav files like signed/unsigned or w64 and bwf?

it seems like it's a random thing.. some normalize to 0.0db some don't

i could force clip the peak sample but that always introduces more minute dc offset..


can i not have zero dc offset and a peak on every sample as 0.0db?

[Meffy]

Yes, you can. If you'd read one of my previous posts you will find a procedure for doing so.

[flux82]

ok tried that but it didnt work.. the top sample still read -0.0db it won't go higher plus when i do pencil it in it changes the file to give dc offset again..

i was hoping it would be some weird file header issue and resaving it somewhere else would sort it out..

i think it must be a wavelab issue of displaying files..

other applications show the file as 0.0db wavelab says its -0.0db

yet wavelab can display 0.0db seemingly randomly depending on what file is in it.

well can anyone recommend me a good offline audio analyzer with dc offset peak and the rest? some digital oscilloscope?

this thread is pointless!! omg im sorry everyone!

[Meffy]

i think it must be a wavelab issue of displaying files..

Yuppers. Nothing worth worrying over. Normalize to -0.3 dB and be part of the happy band of agreeable tonists. Good luck to you.

[soul_junkie]

I must say, for what seemed at the outset to be a fairly odd thread, destined for failure, it has kicked on to provide an interesting discussion with some nice bits of info and opinions.

Well done KVR.

@Meffy I love it when you get the chance to open up the throttle and talk tech, rather than following around idiots & closing threads.

[Meffy]

Thanks, but I'm still not sure whether it makes any sense -- the bit about inverting, specifically, which is the only part that particularly interests me. Too busy to try it. :-} Maybe another day.

[C00kie]

when i do pencil it in it changes the file to give dc offset again..

Like 0.0545% ? Any DC offset less than 1% or so can be ignored.

other applications show the file as 0.0db wavelab says its -0.0db
yet wavelab can display 0.0db seemingly randomly depending on what file is in it.

Could you post a small example file of each case?
If the files are in floating point format, then I don't understand. But in 16bits you have the following phenomena: the sample value can be inbetween -32768 and +32767. If you have basic waveforms without any DC offset, there won't be any sample with the value -32768. But it's a legal value, and it's up to the algorithms to use it or not. Maybe WaveLab shows -0.0 if -32768 is not present and 0.0 if there is.

well can anyone recommend me a good offline audio analyzer with dc offset peak and the rest? some digital oscilloscope?

I use CoolEditPro. Have you tried Wavosaur? What's the other audio editors you're already using?

this thread is pointless!! omg im sorry everyone!

It's not me saying that

[flux82]

yeah i've tried wavosaur to normalize

wavelab gives a read out of the sample as -0.0db still

i think i'm settling on the fact it must be the way wavelab behaves since all other editors seem to display 0.0db the same file as

but here's two samples the first is a crash the second is a clap

the crash displays as 0.0db in wavelab while the clap displays as -0.0db

crash, sample 01 = 0.0db, no dc offset

clap, sample 02 = -0.0db, no dc offset

http://rapidshare.com/files/411385774/2samples.zip