• Welcome to the General Discussion forum for UAD users!

    Please note that this forum is user-run, although we're thrilled to have so much contribution from Drew, Will, and other UA folks!

    Feel free to discuss both UAD and non-UAD related subjects!

    1) Please do not post technical issues here. Please use our UAD Support Forums instead.

    2) Please do not post complaints here. Use the Unrest Forum instead. They have no place in the the General Discussion forum.

    Threads posted in the wrong forum will be moved, so if you don't see your thread here anymore, please look in the correct forum.

    Lastly, please be respectful.

Mythbuster: digital summing, track volumes & master fader

Eurocide

Active Member
As discussed several times in this forum there came up the question whether to keep channel volumes low in order not to clip the master output or just to put down the master fader.

Myth: it sounds better to keep the channels low instead of setting down the master fader.

testing setup:
Cubase SX
2 stereo tracks with sine wave 100 Hz 0dbFS
panning rule \"-3db\" chosen

Test A:
- All channel faders at 0db
- Master output is clipping and it is audible.
- Master fader exactly down to the point until no clipping occurs (-3.0db)
- Export audio the resulting wave

Test B:
- each channel fader -3db
- Master output peak at 0.0dbFS, no clipping
- Master fader at 0.0db.
- Export audio the resulting wave

Wave A minus wave B resulting 0.00000000

test repeated with massive overload of track volumes (+30db) and more tracks: same result.

End of myth, because both ways give exactly the same result.

If somebody please could do this test in Logic & ProTools?
 

Giles117 DP

Active Member
So you are saying that you hear the clipping on test a and it sounds like test B ? which is where you hear no clipping????

Confuzzled.

Nevermind I read it harder you pulled the master down.

However that is with 2 tracks try that test with 24 tracks. and try audio not a sinewave.....

I dont know many people who mix with a sinewave.

I have seen gear that scopes out great sound like crap... But that is just my experience.
 

Giles117 DP

Active Member
Mmm Hmmm. I remember when they taught me that back in 1984 in my digital class......

According to the math digital summing is spot on. But according to our ears.....Now there is the real test...

so again I say do at least 15 to 20 tracks and get back with me.....

remember basic log. 2x = 3db increase, and so on....

so take 24 tracks do your test and give me the results....
 

bugsstar

Member
I'm looking into this summing buss myself.
What i found out and already knew is never start clipping.
If the channel peaks and gives a red light bring it down.
Once you got more tracks over 16 bring down every track.

The reason why many people hate daw mixes in the box is due the fact that they are clipping their tracks (it might not sound bad one channel at a time) but if you ad plug ins and your meter is clipping at unity gane , don't bring down your fader but decrease the plug in amps.

In the end it will offer you a more open and much better mix.

If your channel is clipping at unity gain and so is your master ofcourse you know your're messing up the audio path (digital or analog) so don't bring down the faders because it is not the same.
The headroom is there to mix everything together if you have everything up loud your dynamics in your song will degrade and you will get distortion on the lower parts.

If all the softwares would just give a digital click instead of their clipping fixes , people would have started mixing better years ago.
You might not hear it due to those non digital clicks anymore but you're not getting the best mix and it sound boxy and dull.

Once you add plugins bring everything down.
-18dbFS is a standard calibration for 0db on the analog domain.
With a tape machine you record around 0db and a little above.
So when playing back a digital session , try to work on it like you would on analog.

If you will use analog outboard gear or summing mixers , you should stay for sure around -18dbFS

imo
i'm still reading many articles to back this up completely.

ever wondered why so many ableton mixes are bad ? because many of those users are or were DJS and if there is 1 group of audio users who clip than it's the DJ's. :)
 

MRoadster

New Member
If the masterbus is overloading and there is audible clipping it's only your soundcard that can't handle the signal that's been fed to it. Try the following: Mixdown while overloading on the masterbus. Record in 32 bit floating point format. open the file in an audioeditor eg. Wavelab. You'll see that the file is clipped... now normalize the file and magic happens... all the clippings are gone and you'll have a perfect recording. I never tried it but i wonder if it's exactly the same as a recording done at exactly 0db
 

tapeop

Member
MRoadster said:
If the masterbus is overloading and there is audible clipping it's only your soundcard that can't handle the signal that's been fed to it. Try the following: Mixdown while overloading on the masterbus. Record in 32 bit floating point format. open the file in an audioeditor eg. Wavelab. You'll see that the file is clipped... now normalize the file and magic happens... all the clippings are gone and you'll have a perfect recording. I never tried it but i wonder if it's exactly the same as a recording done at exactly 0db

This is totally insane and makes no sense at all.
 

polygen

Active Member
what levels to use in order to avoid signal reconstruction errors aka \"inter-sample overs\" (and wtf that means, anyway)
http://recforums.prosoundweb.com/index. ... 4918/0/0/0

this thread starts being really interesting around pages 6-7, when bob katz ( http://www.digido.com ) and paul frindle (sony oxford engineer) start \"taking it over\"...but i recommend reading the whole thread nevertheless.
 

bugsstar

Member
MRoadster wrote:
If the masterbus is overloading and there is audible clipping it's only your soundcard that can't handle the signal that's been fed to it. Try the following: Mixdown while overloading on the masterbus. Record in 32 bit floating point format. open the file in an audioeditor eg. Wavelab. You'll see that the file is clipped... now normalize the file and magic happens... all the clippings are gone and you'll have a perfect recording. I never tried it but i wonder if it's exactly the same as a recording done at exactly 0db


-This is totally insane and makes no sense at all.-


i completely back up that statement as being in no sense at all and weird

I can add any other soundcard and the file will still distort.
I even have mastered commercial cds that are distorting due such handlings that there might be more than 0dbFS.
Look at the waveform of for example weezer's latest album : hardly any peak in the waveform all cut off !

Once a digital file hits 0dbFS in 16 bit 24 bit 32 bit or even 128 bit
0dbFS = 1111111111111111111111111111111111111111
there is nothing above that level !!

only thing that such a software can do is apply roundings on the clippings
but your file is still distorted and worse than any other recording method.

i'm still figuring out what the 32bit buss actually does.
as i find no technical information about that issue.
But i would never decrease or even increase the master fader unless it chooses you how to use the 32bits but that's what i need to find out.
 

bugsstar

Member
thanks polygen

i'll start reading it now and hope to find solutions how the summing works
 

bugsstar

Member
\"Plec wrote:
Haha!! I've never seen so many irrelevant posts in one spot before It's all math and theory... stick with it!

Eurocide wrote:
You've made my day! \"


that's easy to say
not understanding what is going on with a certain piece of gear you have.
atleast i don't have to hear your stuff because it will show
and another reason to stay away from hobby-ist forums as there is no discussion possible into finding out what is really going

actually he didn't make your day as you just don't understand why he says it's all math and theory

you think he's proving your point to go and clip your channel faders and clip everything and just pull down the master fader to come up wth your mixes
 

Eurocide

Active Member
bugsstar said:
actually he didn't make your day as you just don't understand why he says it's all math and theory

you think he's proving your point to go and clip your channel faders and clip everything and just pull down the master fader to come up wth your mixes
OK, misunderstanding approved.
But why are you presuming that I do clip my channels in my productions?
If you doubt 20 years of mixing experience I am offended.:evil:
I do understand what's going on. My test works in Cubase SX. But I don't know, how other audio programs behave. That is still to be tested.

bugsstar said:
and another reason to stay away from hobby-ist forums as there is no discussion possible into finding out what is really going
so then, oh grand master, enlighten me because I am not a hobby-ist and have to feed my family with the money earned with my "ugly clipped productions" as you suspect.
 

bugsstar

Member
i do hope you don't clip your recordings at recording and i do know you don't but we're trying to find out at mixing .

but even at mixing i'm trying to understand how a 32bit internal buss works.
because at the end it it still 24bit and you will lose bits at the lower levels because it will be cut off when put back to 24bit.

i can understand why 32bit simplisticly speaking, because you will not overload when all the tracks are summed. But if that goes louder and for example you're working with classical music , you'll lose some of the lower volumes when it goes back to the final 24 bit.

Your test at the beginning of the discussion 2 waves at 0dbFS if it clips for example +6db and i pull down the fader 6db i still hear the clipping I need to pull it down more than 6db. So It's not that simple.

do the test with pink noise or white noise. There are many factors and it' s not just getting rid of the myth analogy when nothing is explained.
As i do hear a difference when everything's is mixed too loud on a DAW.

20 years so you've used many analog consoles.
an analog console has headroom.
0dbu is -18dbFS on your protools or whatever daw fader you have.
So you know about headroom or when a certain stereobuss would not sound good at a certain time.
So 0dbFS you are 18db higher than an analog 0db meter, depends if it is VU so it is a bit lower...

i would never want to clip an analog channel fader, unless if it adds a certain sound or feel and such machine does have headroom. a daw doesn't if you mix at 0dbfs. But i would never clip an analog channel +18dbfs.. i do't think i ever did that not when it still sounds good.

If everything becomes 32bit ... 0dbfs at 32 bit is the same as 24 or 16 or 8. That's what i've alwayes learned. The extra bits at bits at the lower registry so lower volums (reverb tails etc..)

So if we still think that there is a higher bit number than 0dbfs , we are wrong i think because there is no more db in digital above 0dbfs

for example the list at the bottom
http://www.joemeek.com/faq.html

so recording a track i always stay between -20 and -8 dbfs.
With mixing i would even put it lower sometimes because no noise will be added. and you have 24bit dynamic range, which is alot.

I want to know what the 32bit internal buss does.
because if you look at your masterchannel and pre-fader you'll see it clipping the internal buss so i want to know what the fader does when lowering it.
It will only not let you clip the D/A convertor but i think it still is clipping the internal buss and changing the sound.

i'm not trying to offend someone
just to get reactions and explanations
and concerning issues like this it can only be gotten when talking on the edge :)
especially when a simple sentence like : Haha!! I've never seen so many irrelevant posts in one spot before It's all math and theory... stick with it!

i don't think there will be a difference between cubase sx (on pc ??) and logic.
many tests have happened concerning this, which are myths that a program does sound better than this or that.
It can on a certain level but i believe that has to do with how the metering in the program works.

i have apogee converters, clocks , amek preamps, ect... i can sound the same as any PT HD, LE, nuendo any system out there when with just plain mixes (no plug ins, or the same plugins) ofcourse if those daws have the same converter setup as i'm.
 

bugsstar

Member
one thing i do want to make clear

Eurocide, you are right
regarding:
your setup with 2 signals and that those can not clip the masterbus.
The masterbus unlike an analog desk controls the ouput to the DAC.

now the fact about the channel faders.
I'm talking about clipping those tracks.
That's my main concern of investigating.

reason for this is i've seen many people say you can clip (go in the red) with a channel fader and ofcourse never with a master fader as it just is a fader for your DAC ouput (in other words).

If i look at a datapath of a protools system with 48bit bus.
It appears that the channel fader operates at 24bit.
So if you add plugins and you are clipping the channel led you are actually clipping.

That's the problem i have with the statement of an audio file where you can see the clipping and suddenly not anymore.

In a 32bit system you can still clip but if you clip and lower the master fader you are still clipping . 0dbFS stays 0dbfs and the above bits get thrown away.
theoreticaly this is possible... if you have to lower the master fader for example -46db i think it actually clips the internal 32bit bus. could be lower... you're not clipping your DAC at that rate and are using the full bit range of that DAC that is connected.

I must say that a sine wave at 0dbfs and adding another one isn't clipping the master bus and lowering that master fader will fix the DAC clipping but the internal bus isn't clipping. That is correct !

But i do believe both channel faders are clipping although it doesn't show but i do believe frequencies get cut off.

All what i'm saying is never to clip the channel fader !
If you have a normal mix , no plug ins and peaks are around -3 . That mix will be the same as a -20db mix.

But once you start adding plugins the problems occur as i do believe the pre-fader metering and i watch closely that i'm not going above -6db, as not every peak is shown.

So this confusion has all started on myself and others with your statement of 0dbFS !

if you stated peaks at -6dbFS and 92 channels we would have probably agreed with you in the first place.

I truly believe the channel fader is 24bit, plugins might use 32bit processing but get truncated right away after the plugin so the channel fader is a 24bit channel.
I would never let a channel fader clip and that's the asumption you gave that by just lowering the master fader everythings fine.
I don't believe that and my ears don't believe that.

cheers
 

Eurocide

Active Member
Thanks bugsstar for this long explanation.
I totally agree with you.

My concern was just that there is a lot of esoteric voodoo put into DAW algorhythms where simple mathematics can blow this voodoo stuff away.

One thing, that I found very interesting while my test is this:
the wave file itself has got max. 0dbFS (= no clipping within the audio file!) and when I push the Cubase fader above 0db (let's say +6db) or amplify it with a plugin there is no clipping in the channel itself - as you stated. The audible clipping occurs at the master stage only.
If the channel itself would clip, there would be an audible clipping even if you put down the master fader.
And this is what me brought to the conclusion of taking away a \"myth\".

sorry for any embarrassment. 8)
 

polygen

Active Member
bugsstar said:
I truly believe the channel fader is 24bit, plugins might use 32bit processing but get truncated right away after the plugin so the channel fader is a 24bit channel.
for fixed point systems like protools or saw this is true, native systems will stay 32 bit right untill the end (output "fader" needs to convert to fixed point).

read the link posted, it's all in there.
 

bugsstar

Member
only thing i would still warn is getting any meter peak above -3dbFS.
Reason for this is that not all peaks are told by the peakled.
The lower you stay the more the actually peaks due get thru and not get clipped off.

Polygen , what link ?
that long discussion at the proaudio forum ?

---------------------
>bugsstar wrote:
>I truly believe the channel fader is 24bit, plugins might use 32bit >processing but get truncated right away after the plugin so the channel >fader is a 24bit channel.
>
>Polygen wrote:
>for fixed point systems like protools or saw this is true, native systems >will stay 32 bit right untill the end (output \"fader\" needs to convert to >fixed point).
>
>read the link posted, it's all in there.
----------------------

Now back to the channel strip.
Because that's where i want to take this discussion to if you don't mind.

Polygen you say when the 32bit end but i want to know when does it start.

So if i read that protools 48-bit white paper , there is a little drawing at the beginning (page 2). If i look at it and read the explanation on the following pages it tends me to believe the following path.

ADC (24bit) -> channel strip (24bit) -> plug ins (24bit in, double processing , 24bit out) -> channel strip (24bit) -> other plugin for example (24bit in, double processing , 24bit out) -> channel strip (24bit) -> CHANNEL FADER (this goes into 32bit)

So meaning : if you have it on pre-fader and it is clipping, that is bad and lowering the fader won't help as it doesn't work like the master fader does.

That's why many people including myself mix with pre-fader and keep everything low because it does sound better as you never clip the 24bit file just before it hits the channel fader.

if it doesn't clip and you higher your channel fader in which it does start clipping at for example +6dbfs that's not bad because it is at the point when it turns into 32bit floating ?

is this statement i just made correct ?

now 1 reason that it wouldn't be correct is: when freezing a channel in logic which means it is before the fader, that file is 32bit ! so that would mean that the audio file directly goes into the 32bit floating system.
And would also mean that there is a myth when lowering volume , meaning lower than -6dbfs , you'd still get clippings from peaks the meter doesn't see.

and that would mean the guys at macvideopro (which started this all for me) by saying you can clip the channel fader but never the master fader is true. (master fader ofcourse but channel fader?)

i'll do some tests on monday with clipping files
 
UAD Bundle Month
Top