Different meters will light up at different levels. If you load in digital at zero, then it's the same level. You can also import dual mono into jam, you don't have to bounce again. You can lower the output if you want, but it's not necessary. But if you do lower the level, you'll have to dither it. In 93', they were probably clipping to get the level. Plant it's going to care. I'd just leave it.

Michael,

Thanks for the help. I didn't realized I could import the dual mono into jam. That could of saved me some time. I think I'll take your advise and leave it alone. Back to my tracking and mixing.......

thanks
randy

I'm not familiar with ProTools but if you panned one track hard left and the other hard right, it may be increasing the level in the channel to compensate for the lower level in the other (I know Vegas does this). In Vegas, you can change how the pan control pans; not sure about PT.

The reason it only took .1 db atenuation in jam is because the audio was already clipped when you bounced down to stereo so once you brought it down just a hair, it didn't detect it anymore.

But like the last poster said, the best would be to import dual mono instead. That way, neither channel should have any gain (as panning does) at all.

-Joel

Yeah, I forgot to look last night, but Cubase has some kind of setting that has to do with how it treats stereo information or something. Mine defaults to -3db, but maybe PT doesn't have this default. I think it has to do with a left channel peaking at 0 plus a right channel peaking at 0 equals a sum of +3db. So that is why it does the compensation. But I may be way off here; just going on memory and assumption.
If it sounds really bad when you put it on CD (and you will know if it is clipping more than it should), then I would do the adjustment.

blueinkpen,

You hit the nail on the head. As soon as i read the word "pan" in your post i new what was wrong. In the PT session the dual mono audio tracks were each panned to the center. They should have been hard left and right. The fact that i needed to antenuate the master fader 3 db should have been my clue because the two signals were summing on the stereo bus. I gues i got confussed by the .1 db antenuation needed in jam and didn't take the clipping into consideration. Thanks for thinking this through for me.

I did the dual mono import in jam and got a good result and no clip lights even though the level goes up to within tenths of a db of the red line. This is more as I would expect.
Now i have to go back and look at the mp3 files i made using this process 5 years ago. (yikes)

It's so easy to screw up in this world.

thanks again
randy

If you would have created a stereo audio track in ProTools, it would have automatically panned them. I never thought of them being in mono or I would have suggested that in the first place. Sorry, but a GREAT catch by Reggie.

Something to keep in mind is how close you want to be to clipping anyway. I would advise reading the paper about digital distortion here:

MKH 800's

It is a short read and may help you out.

Cheers!
Nika

Nika, the issue of adjusting a fader in a DAW has far more quality consequence than the fact of being very close or less than close to clipping IMO.

Ammitsboel wrote: Nika, the issue of adjusting a fader in a DAW has far more quality consequence than the fact of being very close or less than close to clipping IMO.

Unless the software is severely broken I can't see why this would be the case. What would simply multiplying all data by a fixed coefficient and then dithering off the bottom bits due to the sound - unless the dither was poorly applied?

Nika

Added noise??? that's one reason. It also depends on how much it's clipped. But as he's pointed out, it's fine so why F*ck with it. The more variables, the more potential problems. there are very few engines that don't degrade the audio somehow.

Michael Fossenkemper wrote: Added noise??? that's one reason.

Now wait a second. You drop the signal by, say, 2dB when it's 24 bit data, necessitating dithering it back to 24 bits after the calculation, which adds noise at a level of -141dB FS. Then you dither it to 16 bit and you complain about the noise added at -141dB FS? At what level do you have to monitor to hear noise at -141dB FS?

Something about this claim seems very specious.

Nika

It's a 16 bit file, it's coming from DAT.

And we're editing as 16 bits? Why in the world? We should first importing it into a 24 bit session at minimum. Then we wouldn't have the noise problems you're indicating.

Nika

Nika wrote:
Unless the software is severely broken I can't see why this would be the case. What would simply multiplying all data by a fixed coefficient and then dithering off the bottom bits due to the sound - unless the dither was poorly applied?

I know you can't see it.
I also know why you say as you do.
I'm just saying that I've never heard a fader in a DAW that didn't degrade the sound in some maner.

This is bread and butter for a ME so you should know about this.

Also, asking a novice to level adjust material that he already has got mastered is simply the wrong thing to do. He will be degrading the sound quality for absolutely no reason.

Best Regards

Especially if he liked what he got and didn't want to change it.

Whether or not it's a 24 bit session or a 16 bit session. Dither for a 16 bit file sits at the bottom of the 16 bit file. That you can hear. If it's noise shaped, then you'll hear the effects of that too.

Ammitsboel wrote: I know you can't see it.
I also know why you say as you do.
I'm just saying that I've never heard a fader in a DAW that didn't degrade the sound in some maner.

This sounds like mythology to me. Can you take a 24 bit signal, drop it by 3dB. Raise it by 3dB and hear a difference? Can you see the difference on a null test? How does the difference manifest itself. I'd like to see this on a fixed point system such as Protools TDM.

This is bread and butter for a ME so you should know about this.

What is bread and butter is that many people say this. What is not bread and butter is that it happens.

Also, asking a novice to level adjust material that he already has got mastered is simply the wrong thing to do.

Agreed there.

Nika

Michael Fossenkemper wrote: Whether or not it's a 24 bit session or a 16 bit session. Dither for a 16 bit file sits at the bottom of the 16 bit file. That you can hear. If it's noise shaped, then you'll hear the effects of that too.

Dither for the 16 bit file sits at -96dB FS or so. When I open it into a 24 bit session I now have dither at -96dB FS or so but I also added noise at the -144dB FS level. The overall gain to the noise is +.03dB, from around -96dB FS to around -95.97dB FS.

As for "hearing the effects if it's noise shaped," what do you mean by that? Since gain change is a linear process how will that manifest itself in an audible way?

Nika

If you use a radical noise shaped dither like UV22 or Pow-r #3 or whatever, you can hear it excentuate certain upper freqs. If you have an arsenal of different dithers, pop them in and out and listen to what it does to the audio. It's not as linear as you may think.

As for noise added, depends on the dither and at what freq's. It's also random and not static.

Nika wrote: [quote=Ammitsboel]I know you can't see it.
I also know why you say as you do.
I'm just saying that I've never heard a fader in a DAW that didn't degrade the sound in some maner.

This sounds like mythology to me. Can you take a 24 bit signal, drop it by 3dB. Raise it by 3dB and hear a difference? Can you see the difference on a null test? How does the difference manifest itself. I'd like to see this on a fixed point system such as Protools TDM.

Sure, do the test and see/listen for yourself. Please do it on a decent quality reproduction system.

One of the biggest flaws in the industry was to invent the term "nothing will happen"!

Best Regards

Michael Fossenkemper wrote: If you use a radical noise shaped dither like UV22 or Pow-r #3 or whatever, you can hear it excentuate certain upper freqs.

UV22 is not noiseshaping - just colored dither. POW-r3 is indeed a radical noise shaping algorithm. Both of these accomplish lower in-band noise by pushing much of the noise out of band. I would indeed advise against messing with already-mastered works, but unless the software is broken, a simple gain reduction should also lower the noise of these algorithms and render the audio inaudible. Certainly when bouncing it back to 16 bit we can get compounded errrors from multiple uses of noiseshaping or colored dither, but I cannot see how listening to it in the 24 bit realm would reveal this.

The bottom line that you and I agree on is that he should never futz with already-mastered material. It renders the noise-shaping or colored dither useless and actually erodes the dither by using multiple passes of it. That we agree on.

Nika

Ammitsboel wrote: Sure, do the test and see/listen for yourself. Please do it on a decent quality reproduction system.

I have. As I recall (going back a couple years) the difference manifested itself as mere level-change due - presumably - to different coefficients used for gain as opposed to gain reduction. As I recall, I heard no distortion at all. With a floating point system I suppose I might hear some distortion depending on how dither is implemented?

Nika

Nika wrote: With a floating point system I suppose I might hear some distortion depending on how dither is implemented?

What is it that seems to be the problem with floating point?

Ammitsboel wrote: [quote=Nika]With a floating point system I suppose I might hear some distortion depending on how dither is implemented?

What is it that seems to be the problem with floating point?

There are a few issues depending on how it's implemented, the most significant of which is the inability to properly dither it.

Nika

Could you specify that more in detail?

Ammitsboel wrote: Could you specify that more in detail?

Yes:

MKH 800's

Let me know if you have any questions. This could indeed be accounting for frustration you have with faders in certain programs - but not all.

Cheers!
Nika

Here is a test i did today:

Original 24bit/44.1 file gets attenuated -3db in Sequoia, plays out in a digital loop and gets recorded in track 2 in Sequoia. I then boost track 2 +3db and inverts the phase to compare with the original file.

What i got was absolute nothing, Sequoia couldn't even find anything when i asked it to normalize.

Then i did a listening test and listened carefully to both files, and there was a difference. The difference was that the original file sounded clearer.

I then desided to do the test with 24bit fixed point as the file format instead of 32bit float, but i got the same result.
As i recall Sequoia doesn't dither on other outputs than the main. I have to find out for sure though.

Best Regards

Ammitsboel,

That is an extraordinarily bold claim! You are saying that, in essence, these two files are numerically identical - that when you subtract the numerical information on one from the other you get zero. That would mean that they are numerically identical. Yet you say that one sounds different from the other?

Nika

Nika wrote:
That is an extraordinarily bold claim! You are saying that, in essence, these two files are numerically identical - that when you subtract the numerical information on one from the other you get zero. That would mean that they are numerically identical. Yet you say that one sounds different from the other?

If the meters in Sequoia tells the truth then yes.
Is a program that can compare 2 files just by looking at the numbers in them?

Ammitsboel wrote: If the meters in Sequoia tells the truth then yes.

You said that you normalized the result and there was nothing to normalize. This means that when you subtracted the data in A from B you got zero. That means that A-B=0, so A=B. And you are claiming that you can therefore hear the difference between numerically identical files. That is indeed a fantastically bold claim!

Nika

I'm thinking about this for a second. If 1 file is taken and lowered and printed, then raised and compared to the first. How can they be identical? Isn't it mathmatically impossible for it to be identical? I mean you are tossing out some bits and then adding blank ones. Maybe there is something inherintly flawed with the null test?

here is another question. wouldn't it be possible to create a plugin or harware box that can compare two signals and throw out or filter out the differences? Say for instance you take a signal and raise it by 6db, clipping it in the process or dynamically altering it as with a limiter. This results in harmonic distortion that is added to the signal whether real or aliased. You then take the original signal and offset it by 6db in calculation and compare it to the harmonically enhanced signal and filter out the new harmonics? I know there is a box that will filter out any content above a selected frequency so that aliasing doesn't get introduced, but what about harmonic content that sits within the allowed spectrum. Could just a null scenario work with this or is it more complex than that.

I'm guessing that since the waveform has been altered, that a simple nulling would not only get rid of the harmonic content but would also get rid of any altered waveforms after the point that they have been altered. Am I missing something important here?

I've found out that I have to be more carefull about the digital chain than I had expected, it is not enough just to lay off the ears and do a null test... actually this proves that the null test is totally flawed.

Thangs Nika, you are indeed very open minded towards this! Just what we need to get digital technology better!!

Best Regards

I did the -3dB test in ProTools TDM 6.7 and it nulled to my ears/meters. BUT here's the visual truth (at least for ProTools)... http://www.mixandmaster.com/images/nulltest.tiff

The top is the sum of the original file mixed with the inverted -3dB then +3dB (using faders) file. The waveforms are blown as far up as possible, and there is something there with the top one, even if I can't hear it. The bottom file is an "unassigned bus" recorded. I'll leave it up to you guys to figure out what it all means, but there IS something going on.

Michael Fossenkemper wrote: I'm thinking about this for a second. If 1 file is taken and lowered and printed, then raised and compared to the first. How can they be identical? Isn't it mathmatically impossible for it to be identical?

Floating point can do this. Fixed point cannot.

Nika

Ammitsboel wrote: I've found out that I have to be more carefull about the digital chain than I had expected, it is not enough just to lay off the ears and do a null test... actually this proves that the null test is totally flawed.

Thangs Nika, you are indeed very open minded towards this! Just what we need to get digital technology better!!

Best Regards

Ammitsboel,

Try taking a track and putting it in loop-playback, thereby essentially just playing numerically identical material over and over again. Can you hear a difference there?

Nika

mixandmaster wrote: I did the -3dB test in ProTools TDM 6.7 and it nulled to my ears/meters. BUT here's the visual truth (at least for ProTools)... http://www.mixandmaster.com/images/nulltest.tiff

That's a combination of two things. In Protools TDM it is possible that +3dB and -3dB are not the same numbers. One could be 3.00000001dB, for example, and the difference will manifest itself as a bit of difference (literally). The other is that in Protools TDM you have dither in use after each pass. The difference you are seeing is dither. Try normalizing that track and listening to it. If you hear noise only then it is dither. If you hear noise and some sort of variation of the signal then it's probably the difference between + and -.

Cheers!
Nika