Methods- Marginally Off Topic

"How do you do a digital volume adjustment on a file without math?"

This is the beauty of floating point math. In the 32 bits ~25 of them are dedicated to precision, the rest quantify magnitude. Since 24 bit audio only has 24 bits of precision, in theory a 32 bit float should preserve the wave form at any volume.

Since sound cards don't accept 32 bit float streams (do they?), the wave is dithered on output. If you drop al your track faders ndb and boost the master bus by the same amount you shouldn't lose any detail. I beleive a bunch of people did a loopback test a few months ago using different software packages to see which used float internally, most did.
 
M.Brane said:
Another advantage to running FX on an aux. track is if you want to use the same effect on more than one track, but not the whole mix. Simply bus sends to the aux. and blend it in. I bus my drums pre-fader to a single stereo track to simplify the mixing process and save on system resources.
Click to expand...

I do this too. It is an excellent way to keep CPU usage to a minimum, and if there are several instruments that are supposed to sound like their in the same general place on the sound stage anyway, then why not group their effects (especially verb) up. With a big production type song like my tune "Chrysalis" there are so many tracks that even when using aux busses, my 1.7 Ghz Pentium is about to die, but not without punching me in the face first. I don't think it would even play back if I provided each track with its own reverb.
 
Doug H said:
"How do you do a digital volume adjustment on a file without math?"

This is the beauty of floating point math. In the 32 bits ~25 of them are dedicated to precision, the rest quantify magnitude. Since 24 bit audio only has 24 bits of precision, in theory a 32 bit float should preserve the wave form at any volume.

Since sound cards don't accept 32 bit float streams (do they?), the wave is dithered on output. If you drop al your track faders ndb and boost the master bus by the same amount you shouldn't lose any detail. I beleive a bunch of people did a loopback test a few months ago using different software packages to see which used float internally, most did.
Click to expand...

Interesting. So what your saying is if my software mixer uses 32bit float, I drop my track fader 6db, raise my master 6db, and do a bounce to the same bit rate as my original file that they will null when combined 180 deg. out of phase?

:D
 
Cool!

Thanks for that link. I missed that thread somehow.

Now for the next question: What happens when you use multiple tracks of different program material with differing fader adjustments simultaniously?

What do you do with the master fader?

That test is flawed.

:eek:
 
But how flawed is the sound if you proceed on the basis of the test that Doug cited?
 
In your final mix down, tracks mixed at levels other than what they are recorded at are going to lose some information, unless you are mixing down to float or some other really high bitrate.

All you can do is distribute the error across the whole mix.

The idea of tracking at the level you are going to mix at has some merit in this regard, if the final mix is the same bitrate as the tracks. But since panning adjusts the wave amplitudes fed to the left and right channels, you would have to record in stereo with the track already panned right where you want it. Once a track is adusted bit tho, I'm pretty sure it would suffer the same magnitude of rounding errors that would come with mixing it any old way.
 
Huh? There's math involved even in panning?

So, have I got this right: every change you make in levels, EQ and panning involves some change in the math?

Next question: if you're operating in 32-bit floating point, do those changes in the math make much discernible, HEARABLE differnence?
 
"So, have I got this right: every change you make in levels, EQ and panning involves some change in the math?"
Yup. but you should be fine until mixdown if the software uses floats internally. I don't know much about sq tho, I'm pretty sure it completely mangles the wave form.

I'm starting to feel a bit guilty since this thread has gotten a little off topic from even it's off topic topic in the first place. I dunno, have we changed gears?
 
Dobro - Here's what I've discovered works best in CEP. Do EVERYTHING in 32bit...default open files, record, submix, mix...everything.

2. Don't do submixes for drums. If you have 5 tracks for the drums, so what. Group them together and shove them on track 50 if you want...as long as your machine plays, don't submix.

3. don't submix background vocals.

4. don't submix anything, lol.

I'm not saying this is the easiest way in the world to work, but it makes a cumulative difference. You might not hear any difference in the high hat when you go from it being on its own track with a "previewed" reverb on it to being in a drum mix, printed permanently...but doing that requires the high hat track to go through 2 additional stages of processing that it doesn't have to go through...now, when that hat is again mixed down with the guitars and vocals and everything else, that's another process...then maybe you compress the whole mix, then maybe you add a splash of overall reverb....then you limit it...By the time the hat has been through all of that, those 2 additional steps matter...along with the 2 additional steps you put the crash, ride, and snare through (who cares about the kick, lol).

The math may tell me I'm full of shit, but i know that when I keep everything on a seperate track, and do all the FX as "previews" instead of printing them permanently early in the mix, the whole mix ends up sounding better.

Oh, and thanks to this thread, I'll be doing my Effects through busses; it's easier than "locking" tracks in preview mode, takes up less hard drive space too...although it does seem to be a bit more taxing on the old CPU.
 
Oh, and it's a lot worse in CEP if you work in 16 bit, b/c then each and every process involves 2 additional conversions (from 16 bit to 32 bit, then back to 16 bit again)...it does this invisibly, and I killed a bunch of fidelity for a long time doing this. It's funny, but the first song I posted in the clinic after I started working in 32 float, most of the responses were like..."YOU GOT NEW GEAR!!"...lol.
 
dobro said:
Huh? There's math involved even in panning?

So, have I got this right: every change you make in levels, EQ and panning involves some change in the math?

Next question: if you're operating in 32-bit floating point, do those changes in the math make much discernible, HEARABLE differnence?
Click to expand...

NO! If you're in 32-bit FP (as dobro, chris, & all the other Cool Edit dudes here are), volume and pan adjustments AREN'T GOING TO AFFECT A DAMN THING. I'm no big sound guru, but I do know that much.

I don't mean to yell - I just had too much coffee. And this whole "digital summing causing sound deterioration" thing that I see on every audio forum web-wide is just getting a little bit nuts.

These analog people are infecting your mind!:)

Chris
(who has nothing against analog people - I wish *I* had half a million dollars to spend on making my mixes a little bit "warmer"...:)
 
You're partially right...it doesn't effect anything until you actually mix the track down, but even if it's just to mix from a mono track panned +35 to a stereo track, then yes, it makes a difference. But you're right in that turning a slider up or down or a pan left or right in the multitrack does not effect the original wav file.
 
chrisharris said:
You're partially right...it doesn't effect anything until you actually mix the track down, but even if it's just to mix from a mono track panned +35 to a stereo track, then yes, it makes a difference. But you're right in that turning a slider up or down or a pan left or right in the multitrack does not effect the original wav file.
Click to expand...

Okay, maybe I'm just a bonehead, but I'm not sure I follow you. Obviously, mixing a panned mono track down to a stereo track will result in some changes to the waveform. But the misconception seems to be that when mixing a bunch of files down (and we're talking 32 bit FP files, here) it's somehow better to record low so you don't have to make adjustments with the master fader and therefore incur some summing-related loss - which is incorrect.

Of course any pan, volume, etc. changes will cause just that: changes. But not any math-induced *loss* (as long as your files are consistant with the internal floating point mixing format).

Right?
Chris
 
I'm glad I record everything at 8 bits 11025Hz...in mono!

:D

Excuse me guys....couldn't resist.....I'll try to add some more useful information to this thread next week.
 
chrisharris said:
Oh, and it's a lot worse in CEP .....................................................

after I started working in 32 float, most of the responses were like..."YOU GOT NEW GEAR!!"...lol.
Click to expand...


Chris.....

WTF is 32 float???:confused: :confused:

I record all my tracks at 16bit.....am I stupid?? or what??
 
M.Brane said:


Another advantage to running FX on an aux. track is if you want to use the same effect on more than one track, but not the whole mix. Simply bus sends to the aux. and blend it in. I bus my drums pre-fader to a single stereo track to simplify the mixing process and save on system resources.
Click to expand...

One question/comment about this-I'll often set a verb(or 2) and low roll off eq as my auxes-however any track that is panned and then sent to an aux bus is going to have the return sent down the middle of the stereo image, (unless you pan the aux return)right? So I've been careful to only send stuff panned in the middle to aux sends(Drums bass lead vocs) and add verbs individually to hard panned tracks. In my mind this will preserve your stereo imaging better. Does that make sense?

Good thread by the way...the whole volume thing has been plagueing me. Also if anyone would, what is the average rms volume of your summed mixes? Is there a good and bad or what would be too soft? I sometimes get mixes as close to clipping in cakewalk as possible, with an average rms volume of like -18 or -20....A very weak looking wave-that takes alot of pushing to get to what I would consider a reasonable final volume. Thanks.
 
I didn't notice Lt. Bob made the same joke (almost)...ripping off someone without knowing it, that's really bad. ;) I'm ashamed of myself.

Joro, 32 float is a CoolEditPro term. Well, not exactly. But only CEP users use the word.
 
You must log in or register to reply here.
Back
Top