Good questions.

1 - leaving your soft faders at unity while tracking is a good idea. During mixing, move them all you want.

2 - Each channel's bus will usually allow you some overs (operating at 32 bit float in many softwares) without clipping but you are usually forcing it into your softwares basic brickwall limiter.

3 - Your software's master fader is the *last* piece in the signal chain. That means if you're clipping the main bus and it's forcing your software to use its default limiter - lowering the main fader won't change this. It will still be clipped and limited, but with a lower maximum output. Try feeding horribly clipped samples into your main bus and lowering it by 3dB. The output wave will look flat as a square, but 3dB off of the ceiling.

That all being said - leave the master fader alone. Think of it as a volume knob while you're mixing or recording. (Easier to turn down than the knobs on your monitors...)

As for the track faders, feel free to use them as you would a normal mixer. If you see clipping though...back them off. You may not "hear" the clipping and you may not see it in the master bus, but that's because it's automatically being limited. The only way I'd ever let a limiter touch my stuff is if I turned it on and told it to do so.

Cheers-

J.

Huh? What basic brickwall limiter are you referring to?

If your software uses floating point maths (which most do these days) then no there is no difference beween running two faders at unity and running one at +8dB and the other at -8dB. Unless you consider floating point errors in the multiplications of course, but that is not likely to be significant. There will definitely be no clipping or limiting going on.

You can verify this if your DAW will export 32 bit float files. Import a short audio clip, boost its gain by a ridiculous amount like 50dB so you distort the mix bus horrendously (turn down your monitors first!) then export a 32 bit file. Re-import that file, and turn the gain down by the same amount you boosted it before: notice that the sound cleans up again, rather than just ending up quiet and distorted. ie: that distortion you were hearing was only occuring when the audio was converted to 24 bit integer to feed the soundcard drivers, and the 'clipped' peaks are still intact so long as the audio stays at 32 bit float.

That said, I still mix with my master fader at unity all the time. This is partly out of habit as thats how I work with analog desks, partly because my DAW only allows mix bus plugs to be pre-fader and I like to use metering plugs there, and partly because all bets are off when plugs enter the equation: your plugs might clip overs even if your host does not (depends on how the plug is coded)

IIRs wrote: Huh? What basic brickwall limiter are you referring to?

Try bouncing the clipped files you're talking about down to 16 bit.

They're limited. The software does this (in most cases) to avoid the clipping.

Yes, if you remain at 32 float, this won't occur. But, the last time I checked, Red Book doesn't allow for 32 bit float CDs. You're going to have to cope with the amplitude at the 16 bit level at some point, it may as well be in the initial mix phase.

I'm not aware of any DAW software that limits rather than clips, unless you specifically tell it to. :?

CEP does - at least it used to. I don't know about Audition.

Cucco wrote: CEP does .

You sure about that? I find that hard to believe. Soundforge doesn't. Neither does Cubase, I'm pretty sure Logic doesn't, and Tracktion definitely doesn't.

Yup.

It's been a LONG time since I've intentionally or even accidentally driven the mix into the digital bus so hard as to cause this, but one specific (and tragic) incident comes to mind and CEP was kind enough to avoid square wave syndrom for me without the use of an add-in brick wall.

In any case, even without the brick wall (which I'm inclined to believe your statements), driving too hard into the digital mix bus is a bad idea. The master fader still resides as the last piece in the routing chain (some software will allow you to adjust this, but I can't think of any scenario where this would help in this instance).

Essentially, it would still clip, it would just do so at -3dB if the fader were dropped by as much (clipping being defined as any 3 consecutive samples at -0dBFS - or 2 for that matter depending upon which definition of digital clipping you subscribe).

No it doesn't clip. Floating point audio provides effectively limitless headroom: the peaks are still intact, even if they clip when they hit your soundcard.

We're still talking about different things here. You're assuming the product is staying at 32 float. I'm talking about when it is bounced (the inevitable outcome of all current 32bit float projects).

When bounced, it will clip. Period. (Again, clipping being 3 sequential 0dBFS samples)

In theory, what you're saying is 100% accurate (or darn near it since 32 float can indeed be clipped, but only with such extremes), however, in practice, it is NEVER a good idea to overload the mix bus for any reason whatsoever.

Cucco wrote: You're assuming the product is staying at 32 float.

I'm not assuming anything. The question was about two faders in a DAW app, and every DAW I know of uses floats when mixing. So its basically impossible to clip, and mixing with the master fader at -20 or wherever is perfectly valid assuming you know that all your plugs are well behaved in this regard. The outcome (when bounced to 16 bit) will be exactly the same as if you mixed with the master fader at unity and all your channels 20dB lower.

This is simply incorrect.

Yes, when mixing an audio file in common, modern DAWs, the mix bus stays at 32 float and thus anything mixed, even past "clipping" retains its resolution.

However, when you bounce this wave down to 16 bit, the bus does not automatically normalize so that your loudest peaks are now at 0.0 dBFS. If it's past 0dBFS on a 16 bit scale, it's a clip. Period.

Your example of cranking up the faders and mixing past sanity, then bouncing down and then reducing the amplitude and having your resolution is NOT true if you bounce down to a resolution other than 32 bit float.

I just did it 2 minutes ago and guess what - a clipped square wave of a file.

mixing with the master fader at -20 or wherever is perfectly valid assuming you know that all your plugs are well behaved in this regard. The outcome (when bounced to 16 bit) will be exactly the same as if you mixed with the master fader at unity and all your channels 20dB lower.

This is simply not true.

Again, only if you remain at 32 bit float. If you bounce to 16 bit (which you will), and the mix bus is clipped, you will not regain your resolution.

I don't care what resolution the mix bus is at, when you bounce down to 16 bit, your audio will be ruined - period.

Also, the Master fader is AFTER the mix bus in the routing in all current DAWs. The summing and bouncing occurs AT the mix bus. This means if you've clipped the mix bus (according to the 16 bit file that you're rendering) and then the master fader is down 20dB, you'll have a clipped wave which happens to have 20dB of headroom above the clips.

Have you actually tried what you're talking about?

If what you're saying were true, my mastering work would be TONS easier since most of the clipping I get is NOT from misuse of limiters, it's from people mixing too hard into the DAWs summing bus.

Yes of course I have tried it. Have you? You are dead wrong.

The audio streams are all mixed using floats, then scaled by the master fader, again as floats, and only then will they be truncated to 16 bit when you bounce. ALL the math is done using floats (with practically infinite headroom) so as long as you stay below 0dBFS at the output your 16-bit bounce will be just fine. Mathematically it does not matter whether you scale the individual tracks before summing, or scale the mix bus after summing, so long as all those processes use floating point math (which they do).

The proof is going to have to be in the pudding.

As soon as I'm free this weekend, I'll upload 3 graphics.

Graphic 1 - Control. This will be a 2 track file which is bounced from a multi-track where there is no clipping what so ever (or exceeding the 0.0dBFS mark)

Graphic 2 - Pushed Track Faders. This will be the 2-track graphic from a multi-track file whose faders have been pushed past the point of clipping and then bounced to 2 track.

Graphic 3 - Pushed Track Faders - Lowered Master Fader. This will be the graphic of a 2-track bounced file from where the track faders were pushed past bus clipping and then the master fader was lowered to compensate.

If that doesn't clarify, I don't know what will.

Okay...before I packed up the rig to take it on remote, I did this real quick.

Picture 1 - Regular file - no clipping - plenty of head room

(Dead Link Removed)

Picture 2 - Same exact clip boosted 20 dB using soft faders and bounced to 16 bit - notice all the square tops

(Dead Link Removed)

Picture 3 - Same as above with the master fader lowered 20 dB and then after bouncing, normalized. Notice how similar it is to picture 2.

(Dead Link Removed)

By the way, I put both 2 and 3 into Sequoia again and tried to lower the gain to regain my resolution...nothing but clipped peaks.

My point is, the master fader has no purpose in being used to treat the entire audio file, it makes no difference.

Use the channel faders to get the level you want. The master fader (just like on any real console) should really only be used as a volume knob.

I suppose in graphic 3, it would have been helpful if I didn't normalize it. But, now my computer is off and in the process of being packed up.

It looked identical to its current only quieter. Still had clipped peaks all over the place even though the meters showed peaks of only -20dBFS.

Same is the case with my waves 2 and 3 - invert = silence.

Perhaps the routing in Sequoia is more like an analog console. I can, after all, move the master fader to occur before the master bus, which would give me the same results as what you have. However, I cannot possibly think of a reason where this would be a positive thing.

On an analog console, if you clip the channel, that's it. The same is the case with my default routing in Sequoia. Again, this can be changed, but why on earth would I want to?

--EDIT--

Well, thanks to 32 bit, if you clip a channel in Sequoia, that's not it. However, if you clip the master bus, then bounce to 16 bit, that is it.

Cucco wrote:
Perhaps the routing in Sequoia is more like an analog console.

No, it means Sequoia has no headroom at the mix bus, unlike every other modern DAW. Are you running the latest version?

I think there are two different arguments happening here.
So, I did the test as well with a 48k/24bit Vegas session with all fx chains bypassed.
Each result was rendered to 44.1/16bit.

Image 1 - Multitrack session, all faders +12, all files normalised per track, Master at 0.

Image 2 - Same session, same settings, Master at -27.

Image 3 - Results of both.

Cucco wrote:
--EDIT--

Well, thanks to 32 bit, if you clip a channel in Sequoia, that's not it.

so I am correct then, is what you are saying.

Cucco wrote: However, if you clip the master bus, then bounce to 16 bit, that is it.

of course. no-one in this thread has said any different.

Yes - I am.

I've got Cubase, Tracktion and Live right here on my laptop and none of them, if the mix bus is clipped and bounced to 16 bit, will recover anything above the clip.

I keep talking about bouncing to 16 bit, you keep referring to 32 bit. There IS a difference.

Frankly, I'm getting a little tired of arguing since we're CLEARLY talking about two different things.

Yes, if you're mixing and staying at 32 bit float, you can clip the hell out of the mix bus and never lose your resolution.

If you plan to mix to 16 bit and your mix bus is clipped, you will lose your peaks.

Unless, your master fader is set up to attenuate PRIOR to the mix bus and you can end up with a signal which never reaches 0.0 dBFS using the reduction at that fader.

NO digital mixer has headroom at the mix bus. It's maximum input is 0.0 dBFS. Again, this is TRUE when you're working with a 16 bit medium. I don't give 2 shits at what resolution you mix.

If your final product is 16 or even 24 bit and your mix bus is clipping, your peaks will be lost on your final product.

Cucco wrote:
NO digital mixer has headroom at the mix bus.

I just proved that Tracktion has at least 48 dB of headroom at the mix bus.

Cucco wrote: It's maximum input is 0.0 dBFS.

Wrong. 0dBFS is the maximum value that can be sent to your soundcard, or stored in a 16 or 24 bit file, because all these involve truncating the wordlength and converting to an integer format. The mix bus in a properly coded floating point DAW has virtually infinite headroom.

Cucco wrote: If your final product is 16 or even 24 bit and your mix bus is clipping, your peaks will be lost on your final product.

Your peaks will be lost only if the software truncates the wordlength before scaling the volume back down below 0dBFS at the master fader. Tracktion does not and so has bags of headroom at the mix bus. Apparently sequoia does, and therefore has NO headroom at the mix bus. ergo: sequoia sucks. :lol:

Cucco wrote:

I keep talking about bouncing to 16 bit,

why? that was not the question.

the bit depth of the export is irrelevant, so long as you stay below 0dBFS at the DAW output. It does not matter whether you do that with channel faders or master faders however, because (unless you are silly enough to pay $3000 for Sequoia) your mix bus has bags of headroom.

IIRs wrote: the bit depth of the export is irrelevant, so long as you stay below 0dBFS at the DAW output.

Which is exactly what I've been saying. Stay below 0dBFS at mixdown and you're fine.

It does not matter whether you do that with channel faders or master faders however, because (unless you are silly enough to pay $3000 for Sequoia) your mix bus has bags of headroom.

Well, now you're just being an ass about it. Sequoia is perfectly capable of providing this headroom if you route it to do so. I can completely recreate Bent's output if I rearrange the routing so that the master fader is prior to the final bus.

Proper gain staging would tell you to make sure you don't clip your master bus at all, now wouldn't it?

Would you rather have your DAW perform mathematical equations on the summed output of your bus to create lower amplitudes, or would you rather it just to the job of summing and output only? That seems rather simple to me.

BTW- the version of CEP that I referenced was 1.2. Perhaps it was operating at fixed 24 or 16 bit, I don't know for sure since I no longer have it.

Cucco wrote:

Would you rather have your DAW perform mathematical equations on the summed output of your bus to create lower amplitudes, or would you rather it just to the job of summing and output only? That seems rather simple to me.

huh? you lost me there.

Here is what is happening in Tracktion:

1. all the tracks are scaled by their channel faders, at 32 bit float resolution
2. all the channels are summed together, again at 32 bit float resolution (unless you enabled the 64 bit mixing option ;))
3. summed result is scaled by the master fader, still at 32 bit float.
4. scaled result is truncated to 24 or 16 bits to be output to the audio hardware, or stored in a file.

Here is what is apparently happening in Sequoia.

1. all the tracks are scaled by their channel faders, presumably at 32 bit float resolution
2. all the channels are summed together, again presumably at 32 bit float resolution
3. summed result is truncated to destination bit depth.
4. truncated result is scaled by the master fader (presumably at 32bit float resolution again)
5. scaled result is truncated to 24 or 16 bits again to be output to the audio hardware, or stored in a file.

Why the extra step? It just adds another level of quantisation error, and removes all the headroom.

:?

The truncation is only there because I put it there by bouncing to 16 bit. Sequoia stays at 32 float all the way through until I bounce.

I assume Vegas operates at fixed 24.
That being said, 0dB = 0dB, anything over is a clip.
All current bitrates have a 0 point at which overs result in a clip.
I don't care what I'm using, a DM2K, an 01V, Vegas, Protools - they all can and will clip when they go beyond zero (ignoring any built in limiting that the mixers might have)...

If you have one fader at +48 and your master is at -48, that's not headroom, nor does it prove you have headroom in any way, it proves that you would be pushing into clipping if you put the fader at +48.1, or the master at -47.9...

I seriously don't see how an argument about bit depths here can resolve anything. Now, truthfully, I have noticed a slight increase going from 24bit to 16 in the form of about a 1dB increase overall, but you guys aren't arguing that, nor does that have anything to do with a DAW's headroom or clipping, etc.

Nothing from nothing leaves nothing, Lillian!

Does this prove either side of this argument? I don't know. I think it does, but I'll let you decide.

I see where the mixup lies here and as soon as I'm able to type while not driving 70 mph while typing on a PDA, I'll clarify. I just reread and see where I totally went off on a tangent and why we're talking about 2 different things right now.

if you go back and read, you'll see exactly where my brain went astray. There's no mention of bouncing, nor is there any mention of clipping at the time of bouncing.

Sorry - series of brain farts. My only excuse is that I've only slept 8 hours since Monday.

I feel your pain!

it took us until page 3 to realise we were agreeing with each other!

Ain't life grand?

I'm glad you all agree, and I kind of understand what you are talking about.

What I don't get is how you posted the screen shots form the DAWs!!

I appreciate you guys hashing this stuff out. Honestly. It's a good learning tool. I understand everything except for the point about 32-bit clipping:

To begin with, are you talking about recording at 32-bit float, or an application's resolution at mixdown? Are you saying that material recorded at 32 bits that's been pushed +50 dB beyond clipping and saved at the same resolution can be reimported and attenuated into a clean waveform?

taxman wrote: I'm glad you all agree, and I kind of understand what you are talking about.

What I don't get is how you posted the screen shots form the DAWs!!

Ctrl-Prnt Screen

Then

Open your image editor and paste (Ctrl-V)

patrick_like_static wrote: I appreciate you guys hashing this stuff out. Honestly. It's a good learning tool. I understand everything except for the point about 32-bit clipping:

To begin with, are you talking about recording at 32-bit float, or an application's resolution at mixdown? Are you saying that material recorded at 32 bits that's been pushed +50 dB beyond clipping and saved at the same resolution can be reimported and attenuated into a clean waveform?

As long as you bounce down to 32 bit float, then yes.

If you bounce to any other resolution, no - you're screwed.

Best practice - don't clip your master bus under any circumstance.