digital gain staging

[dannygold]

Yes I know you should not clip A/Ds or D/As. Yes I know you should not clip the master fader. Is it OK to clip individual faders? I say yes and I have evidence to support this conclusion.

I created a full scale 1K sine wave (normalized) and placed it on 24 tracks in Pro Tools LE. I placed 5 trim plugins on the tracks each adding 12 dB of gain. I turned the faders all the way up. Obviously the tracks are clipping about as much as a track can (something like 70dBs of clipping). Then I made a master fader and attenuated it to prevent clipping the master bus and bounced. The results was a clean sine. If there was clipping on individual tracks would it not be evident as a squaring off of the sine?

I'd go further and say clipping a PLUGIN is not a problem either (unless the plugin is written to emulate clipping like say BF76). The last trim plugin is being clipped by about 60 dB and does not induce squaring of the sine wave). Why? Because of the floating point math involved (64bit float for LE and most DAWs).

Anyway, it drives me bonkers when people apply analog mixing "rules" to digital mixing. I heard a colleague say they'd fail a student for clipping any output channel in a mix. Yet I have evidence to say this is not necessarily a problem. I know there was a thread going around here about how great it was to trim pre plugins... agiain why? Unless the plugin is designed to emulate analog clipping probably not an issue. If the plug is designed to clip this might be an effective strategy if the plugin lacks it's own input control. But let's be clear about WHY we are trimming plugin inputs and when we need to (only if the plugin emulates analog and lacks a pad basically).

ASAIK the output clip indicator on a track or purely digital floating point plugin acts as an early warning that IF the track/plugin was sent directly to an D/A without master fader attenuation it WOULD clip. Usually not the case if mixing ITB.

Can we stop applying analog mix standards to digital mixing? Or at least try to understand digital gain staging in a floating point world?


[IMG]file:///Users/dannygold2/Desktop/Picture%201.png[/IMG]

[Ethan Winer]

You nailed it Danny. I show similar examples in my AES Audio Myths video. One example overdrives an EQ plug-in by 18 dB, then compensates with a trim plug-in afterward. The result nulls completely with a second track having the same EQ without the 18 dB boost and cut.

--Ethan

[Wayne]

Originally Posted by dannygold

... I'd go further and say clipping a PLUGIN is not a problem either (unless the plugin is written to emulate clipping like say BF76). The last trim plugin is being clipped by about 60 dB and does not induce squaring of the sine wave). Why? Because of the floating point math involved (64bit float for LE and most DAWs).

...Unless the plugin is designed to emulate analog clipping probably not an issue. If the plug is designed to clip this might be an effective strategy if the plugin lacks it's own input control.

ASAIK the output clip indicator on a track or purely digital floating point plugin acts as an early warning that IF the track/plugin was sent directly to an D/A without master fader attenuation it WOULD clip. Usually not the case if mixing ITB.

Can we stop applying analog mix standards to digital mixing? Or at least try to understand digital gain staging in a floating point world? ...

I have a quickie reason for you –and I'm being simply practical here (not trying to be cute or anything..

If you record with a good nominal levels, i.e. -20ish rms, the mix starts off on the right footing, headroom once in, of concern or not, is not an issue. If the tracks are out of whack, I'd fix them to where they are near nominal alignment and move on.
Once this bit is cleaned up' frankly I have no use for screen wasting track meters.

As for track/mix creep' I still prefer to maintain the analog' mode rules ITB simply as a work method/ethic- I try to build with a bit head room on the mix in the roughing in stage (..come in a little under), otherwise if it's coming in a bit high or low I'll use the pre gain/trim on the master (or subs), fine, no problem.
I guess though allot of this to my mind is 'stay nominal, never worry about some 'gotchas or not. I can look at my 'master at zero' and see/hear/know right where it (and my master comp/limiter if applicable) will landing back in 'real world and...
I can shift the whole shebang' (habits.. ) over to an analog' setup and be a happy camper to boot.

[dannygold]

I find the choice of words "work ethic" interesting here... it's not an ethical issue. You can mix how you want... and I get what you're saying about moving between the worlds of analog and digital (if you gain stage the same in both it's easier to transition)... all I'm saying is there's no technical reason to apply the 1970 gain staging to 2010 DAWs.

[j-uk]

Originally Posted by dannygold

Yes I know you should not clip A/Ds or D/As. Yes I know you should not clip the master fader. Is it OK to clip individual faders? I say yes and I have evidence to support this conclusion.

I created a full scale 1K sine wave (normalized) and placed it on 24 tracks in Pro Tools LE. I placed 5 trim plugins on the tracks each adding 12 dB of gain. I turned the faders all the way up. Obviously the tracks are clipping about as much as a track can (something like 70dBs of clipping). Then I made a master fader and attenuated it to prevent clipping the master bus and bounced. The results was a clean sine. If there was clipping on individual tracks would it not be evident as a squaring off of the sine?

I'd go further and say clipping a PLUGIN is not a problem either (unless the plugin is written to emulate clipping like say BF76). The last trim plugin is being clipped by about 60 dB and does not induce squaring of the sine wave). Why? Because of the floating point math involved (64bit float for LE and most DAWs).

Anyway, it drives me bonkers when people apply analog mixing "rules" to digital mixing. I heard a colleague say they'd fail a student for clipping any output channel in a mix. Yet I have evidence to say this is not necessarily a problem. I know there was a thread going around here about how great it was to trim pre plugins... agiain why? Unless the plugin is designed to emulate analog clipping probably not an issue. If the plug is designed to clip this might be an effective strategy if the plugin lacks it's own input control. But let's be clear about WHY we are trimming plugin inputs and when we need to (only if the plugin emulates analog and lacks a pad basically).

ASAIK the output clip indicator on a track or purely digital floating point plugin acts as an early warning that IF the track/plugin was sent directly to an D/A without master fader attenuation it WOULD clip. Usually not the case if mixing ITB.

Can we stop applying analog mix standards to digital mixing? Or at least try to understand digital gain staging in a floating point world?


[IMG]file:///Users/dannygold2/Desktop/Picture%201.png[/IMG]

It still makes sense to keep the levels down, you are at some point leaving the mix bus....

[dannygold]

Originally Posted by j-uk

It still makes sense to keep the levels down, you are at some point leaving the mix bus....

As my screen caps show master fader attenuation prevents clipping when leaving the mix bus... so sure you can keep individual faders down but that's not really better than attenuating the master fader per se. More traditional maybe. And there's other reasons to prevent fader creep (having somewhere to go to turn things up comes to mind)

The point is I wouldn't worry about a red clip indicator on an individual tracks in an ITB box mix... if they were directly assigned to D/As spilling out to a board that'd be a different matter. ... I certainly wouldn't fail a student for this as my colleague said he would... that's just totally arbitrary.... if it's not clipping it's not clipping... oddly enough the clip indicators on a channel really don't seem to indicate clipping... they indicate potential clipping if sent to a D/A/.

Clipping an A/D or D/A is a different matter. I"m not saying that's OK. But in the later case master faders do a great job of preventing this.

Originally Posted by dannygold

Yes I know you should not clip A/Ds or D/As. Yes I know you should not clip the master fader. Is it OK to clip individual faders? I say yes and I have evidence to support this conclusion.

Some plugins (particularly ones which rely on dynamics processing like compressors) rely on correct gain-staging. I'd say most plugins do.
What matters is what comes into and goes out of the plugin - and what plugin.
As for the individual track output going into the summing bus, no, it doesn't matter for the most part.

[ms70]

Originally Posted by dannygold

Yes I know you should not clip A/Ds or D/As. Yes I know you should not clip the master fader. Is it OK to clip individual faders? I say yes and I have evidence to support this conclusion.

if you'll find ever time to read through this huge thread, i bet you will record/mix at -18dbFS from then on! :-)

[Bill@WelcomeHome]

Originally Posted by dannygold

The point is I wouldn't worry about a red clip indicator on an individual tracks in an ITB box mix...

Assuming that all you've said is true, so what? I'm not trying to be smart, I'm trying to discover what you are promoting here. What do we have to gain? (he said gain...yuk, yuk...)

If I follow traditional gain staging as I've done since the 1960s, and all things are equal, when I am done tracking and mixing the end result is going to sound great and go off to mastering with levels anyone can work with. What will be better if I clip every track?

[007]

Ok here we go again...

[dannygold]

Originally Posted by soulstudios

Some plugins (particularly ones which rely on dynamics processing like compressors) rely on correct gain-staging. I'd say most plugins do.
What matters is what comes into and goes out of the plugin - and what plugin.
As for the individual track output going into the summing bus, no, it doesn't matter for the most part.

Agree with that... sure.

[dannygold]

Originally Posted by ms70

if you'll find ever time to read through this huge thread, i bet you will record/mix at -18dbFS from then on! :-)

The entire basis of that thread is you should gain stage a DAW like an SSL which I entirely disagree with (at least in terms of faders... plugins like dynamics or those that purposely emulate analog perhaps as the poster above points out). I mean if you WANT to gain stage a DAW like an SSL it won't hurt, but you don't need to .

Look at the screen caps I posted... if you clipped an individual fader on an SSL by 70dB you'd be in distortion. CLEARLY on a DAW that is not true. I posted evidence of that in the OP.

[j-uk]

Originally Posted by Bill@WelcomeHome

Assuming that all you've said is true, so what? I'm not trying to be smart, I'm trying to discover what you are promoting here. What do we have to gain? (he said gain...yuk, yuk...)

If I follow traditional gain staging as I've done since the 1960s, and all things are equal, when I am done tracking and mixing the end result is going to sound great and go off to mastering with levels anyone can work with. What will be better if I clip every track?

Regardless of anyones conceptions or misconceptions of digital gain staging this advise makes the most sense.
I agree that clipping internally is not an issue but I really don't see the benefit of doing so over a traditional analog gain staging.
One of the apparent problems of too much gain is that even though it's not clipping its impossible to set a compressor to subtle if the signal is above 0dBFS.
To each his own. If you find a benefit in working this way then good for you. I'm happy to know that if a channel is peaking than I don't have to worry too much about it but in general I will always keep the levels down in the box.

[dannygold]

Originally Posted by j-uk

Regardless of anyones conceptions or misconceptions of digital gain staging this advise makes the most sense.
I agree that clipping internally is not an issue but I really don't see the benefit of doing so over a traditional analog gain staging.
One of the apparent problems of too much gain is that even though it's not clipping its impossible to set a compressor to subtle if the signal is above 0dBFS.
To each his own. If you find a benefit in working this way then good for you. I'm happy to know that if a channel is peaking than I don't have to worry too much about it but in general I will always keep the levels down in the box.

Sure fair, enough... I guess all I'm saying is it's kind of silly when people turn all tracks down because one track is "clipping"... this can have major effects on the mix if pre fader sends are being used and is actually not necessary.. THAT is why it's good to understand digital gain staging.

[Timur Born]

I miss some details here:

1) Adding digital gain and then subtracting by the same amount does *not* necessarily null, it's even unlikely because of the rounding that GUIs do for displaying/entering decimal dB value that under the hood are floating-point binary values! I once did some checks with both Logic and Live and both did *not* null after adding and subtracting the same dB values.

Even worse, different digital gain-stages (DAWs, Gain utilities, plugins) may behave very different! Again this is mostly because of the decimal dB to binary FP conversion and how much precision a gain-stage allows to input at all (i.e. Logic tends to round rather rough while Live allows many "hidden" numbers after the decimal point).

But the resulting signal may be well below audible range, or even below measurable range if your meter only goes down to -144 dB.

2) In 32-bit floating-point engines going over 0 dB on any single track/effect/plugin means losing the lowest bits. That is because 0 dB = 1 = full 24-bit precision of 32-bit FP. Any bit you put on top of it you lose at the bottom end.

As an example this means that once you go +6 dB on a track that you lose the -144 to -138 dB information.

With 64-bit floating-point this is no problem, because it offers more than 24-bit precision to begin with.

3) Plugins: While the same principles as above usually apply for the internal signal-path of plugins, too, there is *no* guarantee that specific plugins don't work entirely different.

For example: NI Guitar-Rig 2 (AFAI remember) can be clipped at any internal stage, but *not* after its own internal Master fader. It will clip the output to the DAW no matter how much gain you attenuate after the plugin. It simply does not output any value bigger than 1 (=0 dB).

4) Using analog style mixing in a digital environment is still good practice! It *may* not be necessary, but do you check all your new gain-stages and plugins before any project?

Analog style mixing protects you from possible culprits like misbehaving plugins and gain-stages. It also keeps you from forgetting that you're just running the brick-wall limiter hot on your Master bus (I know you are all lazy b*st*rds ). And last but not least it allows you to easily go back to any analog mixer/world without having to twist your brain-paths.

[j-uk]

Originally Posted by dannygold

Sure fair, enough... I guess all I'm saying is it's kind of silly when people turn all tracks down because one track is "clipping"... this can have major effects on the mix if pre fader sends are being used and is actually not necessary.. THAT is why it's good to understand digital gain staging.

I agree wholeheartedly. I also find it shocking that a teacher would fail a student for not trimming a channel internally. Only if the result were audible would this be the correct measure.

As a side note though. The difference between analog and digital gain staging are difficult for students to understand at first. Effectively you are saying that what applies in the first domain is the opposite in the second with the exception of the main mix bus where the rules of the first domain are applied again......
Yea that didn't make much sense to me either........
I would agree with anyone who in facing the onslaught of mass confusion will revert to simply teach analog gain staging at first and conveniently forget about the non- clipping, clipping world of digital.

[Bob Olhsson]

Originally Posted by Timur

...Using analog style mixing in a digital environment is still good practice! It *may* not be necessary, but do you check all your new gain-stages and plugins before any project?

Analog style mixing protects you from possible culprits like misbehaving plugins and gain-stages...

The idea is to avoid any digital signal processing that is operating outside of its design parameters because there is no easy way to know exactly what those parameters were. A real problem with DSP is that inaudible overloads can come back around and bite you upon subsequent processing of the signal.

[UnderTow]

Originally Posted by dannygold

Yes I know you should not clip A/Ds or D/As. Yes I know you should not clip the master fader. Is it OK to clip individual faders? I say yes and I have evidence to support this conclusion.

That is not clipping. It is just going above zero on a meter. Clipping means something specific and a floating point fader going over 0 dB FS is not clipping the signal.

Anyway, on a more important note, some plugins clip at 0 dB FS. (The Waves Linear phase series is a good example of this). Also there is no guarantee that a floating point plugin will behave properly over 0 dB FS. It might not have been tested that way simply because that is not the usual way to us it. It is very good practise to always stay below 0 dB FS to avoid nasty surprises.

So to conclude: You can get away with going over 0 dB FS in some situations but it is a much better idea to avoid it.

Alistair

[Primativ]

Since I have started applying proper digital gain staging, when I coem to mixdown everything is so much easier. I can add compression, limiting, or whatever, without worrying I am going to clip the mixbus.

I prefer knowing that there is a pattern and practical process to my levelling and so when I send off my tracl to the ME he is going to have lots of headroom to play with to make my mix sound fat...

[Timur Born]

Originally Posted by dannygold

I placed 5 trim plugins on the tracks each adding 12 dB of gain. I turned the faders all the way up.

I once did a similar thing in Logic 8, added and subtracted Gain via several instances of Logic's internal Gain Utility. The result:



The spectrum graph says 0/zero/null, the dB meter says +5.5 dB, the speakers say 0. Which is right?

The conclusion: Don't push things to the edge without *willing* cause. You never know into what kind of culprits (like bugs, or in this case DC) you will run. Stay within the (most tested) parameters and be good with it. What are the most tested and common parameters? The range below 0 dBFS!

[Lagerfeldt]

Guide to Levels in Digital Audio

[synthoid]

there is still quite a bit of signal processing code in the world that was developed on fixed-point signal processors for which 0dBFS was a hard boundary. this kind of code was forced to "clip" or saturate when exceeding that level -- there was no other possibility. sometimes this code was converted to floating point code in such a way that the 0dBFS boundary disappeared, but probably just as often the conversion left some artifacts at the 0dBFS point. Best to stay clear of that signal level unless you are very sure how your plug-ins respond.

Another consideration here is that ITB metering is usually missing or incomplete for signals over 0dBFS, so working up there is asking for problems reading out your signal levels.

-synthoid